July 21 2007 all running PC's in the world...or just in the US...will be on 32bit hardware and OS's...running 32bit applications.
You've already lost that one, im afraid.
Ok, you're points are fair enough but you've got to remember the HL2 specs are flawed since people got HL2 free with 9600s. And if there isnt continual evolvement in the industry you cant drag people kicking and screaming into the future.
HL2 isnt the most graphically demanding game on the planet and PC games take longer to develop than ever now because of the requirements of coding.
How many people vowed never to leave 98 for XP and are now on XP? It will be the same with Vista and Vista needs a serious hardware set. It will force people to upgrade. With the advent of future technology to enable people to do more, simultainiously, you will need hardware to do that. It's not so much ever about running one thing at one time, it's about having the avaliable overhead to do MANY different things and still not experience a slow down.
Originally Posted by SGT Lindy Bindibadgi......"Are we talking about software that was made in 1995 or software that doesnt need to be or wont benefit from being multithreaded?"
Read my original post...."If you look at all of the applications that are currently on the market today for your PC, 95% of them are 32bit and support ONE CORE".
Translation....of all of the software you can purchase right now ..or download or whatever 95% of it is 32bit single core enabled.
Out of interest, of all of the software you refer to, how much of it benefits from more than an Athlon 1000 (Thunderbird core)? The point I am getting at is that 95% of applications out there are not CPU-intensive and therefore you're not going to see a benefit anyway. The applications that are CPU-intensive are being developed with multiple threads in mind.
Quote:
July 21 2007 all running PC's in the world...or just in the US...will be on 32bit hardware and OS's...running 32bit applications.
There are a few people on these forums already running Windows XP Professional x64. I think that many bit-tech readers will be running the 64-bit version of Vista by July 21st next year (if Vista has launched by then - of course, we can only hope... ;)). Basically anyone who owns an x86-64-capable CPU - there's little reason not to, IMO.
Originally Posted by Bindibadgi You've already lost that one, im afraid.
Ok, you're points are fair enough but you've got to remember the HL2 specs are flawed since people got HL2 free with 9600s. And if there isnt continual evolvement in the industry you cant drag people kicking and screaming into the future.
HL2 isnt the most graphically demanding game on the planet and PC games take longer to develop than ever now because of the requirements of coding.
How many people vowed never to leave 98 for XP and are now on XP? It will be the same with Vista and Vista needs a serious hardware set. It will force people to upgrade. With the advent of future technology to enable people to do more, simultainiously, you will need hardware to do that. It's not so much ever about running one thing at one time, it's about having the avaliable overhead to do MANY different things and still not experience a slow down.
I lost that one??????? One year from now you think more people will be running dual core applications on a dual core CPU, on a 64bit OS than are on a 32bit OS with single core CPUs???? That is a lot of serious upgrading for a lot of people .just not going to happen in that short of a time. 3 years from now I will be it will be 50/50. There are some people and companies just giving up Windows 98 now .with lots of companies still on Windows 2000.
Most companies wont move to Vista until at least SP1. 90% of home users wont move until they need a new PC. Early vista sales will come from large corporations moving to vista and places like Best Buy and Dell selling new PC's to normal users who by a large margin make up most of the market. Hard core, overclocking bleeding edge people make up less than >1% of the market.
My company is testing Vista right now, my main workstation is running it. We are getting RC1 in August and MS has told us the Enterprise version of Vista will be ready at the end of October...we will see if they are right.
That said we are going to deploy the 32 bit version on any PC in our environment that has 512megs of ram min, with 1gig being the amount we will purchase new computers with. Our full push to approximately 64,000 users will take about 2 years. Current plans are not to deploy the 64 bit version as almost none of our applications will be 64bit save Office 2007. Right now some of their own business apps wont run on Vista like Virtual Server and Office Communicator and we are running a build past Beta 2. It will be enough for them to get the 32bit versions of these apps working let alone create full blown 64bit versions by November 1.
"There are a few people on these forums already running Windows XP Professional x64. I think that many bit-tech readers will be running the 64-bit version of Vista by July 21st next year (if Vista has launched by then - of course, we can only hope... ). Basically anyone who owns an x86-64-capable CPU - there's little reason not to, IMO"
I run Windows XP 64bit at home and its very solid...but honestly there is nothing I need the 64bit version for. My only 64bit apps are Symantec AV 10.0 corp edition, Microsoft Defender, and Virtual Server 2005 R2. The rest is all 32 bit....single core stuff. Some drivers were a problem (camera) in the begining but most of my hardware is pretty mainstream.
At work we are moving to 64 version of Windows 2003 pretty fast. The main reason is not 64bit apps but the removal of the 4gig of virtual memory limitation. It has helped alot on some of our large file server clusters and with our SMS SQL server. However some applications like Exchange 2003 wont run on it.....but the next version Exchange 2007 will only run on the 64bit version.
It will be very interesting to see which version of Vista 64bit or 32 bit Dell and HP will ship with their mass market PC's. I would think the 64 bit versions since all new PC's will be 64bit by then....but if they are worried about compatibility maybe not????
Originally Posted by aggies11 Multi-threaded is one thing, but I don't think we can say that the games have a balanced load between the cores. 2x2GHz does not perform in-game (CPU wise) relative to a 1x4GHz. There are benefits mind you, but not nearly in the range to say that the games are taking "full advantage" of both cores.
Have you ever monitored your CPU usage while playing a multithreaded game? I have, and let me assure you that the game does indeed use as much CPU time as it requires. In some situations, this means that it goes back to the GPU as being the limiting factor. Load up HL1 some time - I've run it absolutely maxxed out on new hardware, and even on a single-core PC, I don't think it ever topped 60% CPU usage, let alone the full thing. Other parts of the system were the limiting factor.
In a properly multithreaded application, 2x 2GHz cores will perform almost identically to 1x 4GHz. When I moved from a single- to a dual-core system, of about the same clock speeds, my speed in threaded applications was almost exactly doubled.
As to the various 64-bit comments... are you running 4GB or more of RAM? If not, you have very little to gain by moving to a 64-bit OS. 2^32 just so happens to be the number of bytes in 4GB, and it's no coincidence. You can be pretty certain that the limit of addressable memory with a 64-bit system is 16 EB - 16 billion gigabytes (well, disregarding the 1024 vs 1000 thing). I don't think it would be an unreasonable estimate to say that amount is probably on par with all of the RAM on the planet, and almost certainly within the country.
Obviously there are some skeptical users on here. Of those who are - have you USED a SMP machine, be it multicore or multiprocessor? If yes, then I'm sure you have your reasons for thinking what you do. My dual-core machine was my first SMP experience. And quite literally after first boot, I immediately decided that I'd never go back. Everything is smoother in general usage, and CPU-intensive tasks never leave the machine in an unusable state. Heck, I've gamed with a recode running in the background, and barely noticed - sure, my FPS was a tad lower than usual, but it certainly was no slideshow. The only way I can get my computer in a somewhat painful state to use is when I've got Xvid transcoding going on while defragging and ripping apart another DVD. I've had occasions where I'd have loved a quad-core system, but the only times I felt such a way was when I was cranking on my machine like it's never seen before.
Just to further refine my point and clarify things.
I agree with Tim regarding everything you said. Multi-core is certainly not a dead end. There are definite, tangible and immediate benefits to going dual core. Both for specifc applications (FWIW I'd consider "media compression" ~= "media encoding" :p), and also for general computing.
But remember the main topic of the thread, "Why buy 2cores, when 4 cores are comming out in 6months?"
Let me put out the arguments that I believe leads to the conclusion that *most* people (who read bit-tech, not just most people in general) shouldn't wait.
- Not all apps can see benefits to multithreading. Those that do, the performance increases are often not always %100. "The benefits of dual core aren't %100"
- For those that do see benefits to multithreading, those benefits will become even smaller when going from 2->4 cores (than when going from 1->2 cores). "The benefits of quad-core are even less of an improvement".
With the above two points, I'd say it's not worth waiting. That quad core doesn't add *enough* to the table to justify waiting. Obviously there is improvement, it's just not enough to warrant delaying an upgrade. Especially when compared to the jump that is/was Conroe.
Thats my whole point. But there is a definite trend in current opinions that extole the virtues of multi-core/paralellization. It's treated as if 2cores essentially result in one "twice-as-fast" single core. And my contention is that is quite far from the truth. Multi-core has it's benefits, but in specific circumstances which aren't usually in the majority.
Again, if you were to give me a 4core chip, I'd take it without hesitation. It can't hurt, and there are benefits to be realized. But it's not the end all and be all, and more importantly the gains from 2->4 are less than from 1->2
Otherwise we'd all be running massive parallel arrays of p166's in our homes :p
Originally Posted by aggies11 Just to further refine my point and clarify things.
I agree with Tim regarding everything you said. Multi-core is certainly not a dead end. There are definite, tangible and immediate benefits to going dual core. Both for specifc applications (FWIW I'd consider "media compression" ~= "media encoding" :p), and also for general computing.
But remember the main topic of the thread, "Why buy 2cores, when 4 cores are comming out in 6months?"
Let me put out the arguments that I believe leads to the conclusion that *most* people (who read bit-tech, not just most people in general) shouldn't wait.
- Not all apps can see benefits to multithreading. Those that do, the performance increases are often not always %100. "The benefits of dual core aren't %100"
- For those that do see benefits to multithreading, those benefits will become even smaller when going from 2->4 cores (than when going from 1->2 cores). "The benefits of quad-core are even less of an improvement".
With the above two points, I'd say it's not worth waiting. That quad core doesn't add *enough* to the table to justify waiting. Obviously there is improvement, it's just not enough to warrant delaying an upgrade. Especially when compared to the jump that is/was Conroe.
Thats my whole point. But there is a definite trend in current opinions that extole the virtues of multi-core/paralellization. It's treated as if 2cores essentially result in one "twice-as-fast" single core. And my contention is that is quite far from the truth. Multi-core has it's benefits, but in specific circumstances which aren't usually in the majority.
Again, if you were to give me a 4core chip, I'd take it without hesitation. It can't hurt, and there are benefits to be realized. But it's not the end all and be all, and more importantly the gains from 2->4 are less than from 1->2
Otherwise we'd all be running massive parallel arrays of p166's in our homes :p
Aggies
I fully agree with you. I can tell when I am doing alot of work on box that is single core....like my laptop with a Pentium M...compared to my work desktop box that has hyperthreading. Bouncing back and forth between 5-7 applications cutting and pasting there are less pauses with the Hyperthreaded CPU.....and even slightly better with my Pentium D at home.
That said I cant run a Virus Scan and play BF2 at the same time with out serious performance hits in both...even with my Pentium D 930 because niether of those applications are dual core aware.....XP will try to spread the threads across both cores but it does not work very well.
If I were playing just BF2 and not doing anything else....having a single core or dual core would not matter.....and that is true for 95% of applications out there.
If I were building a new PC for a friend today...I would go for the lowest Conroe that has 4meg of cache, E6400?? Because I do know that dual core gives you some advantages...and just makes things smoother overall. If this person was my mother-n-law or someone that just surfs the web, uses email, downloads pictures and uses Quicken....then I would go for the cheapest conroe they will make...even a single core celeron version of they are going to make/label one. That type of person is probably moving off of some Pentium III or P4 Celeron anyway and even the slowest conroe would be a huge boost.
There is no way I would recomend a 4 core to anyone unless.....it was cheaper than a dual core and used the same power. These days for me and most of the people I know its all about quiet PC's....quiet, cool and power friendly. I am going to get the low end 4meg conroe to replace my Pentium D so I can turn my Zalman all the way down to the slowest....quietest speed.
Originally Posted by SGT Lindy I lost that one??????? One year from now you think more people will be running dual core applications on a dual core CPU, on a 64bit OS than are on a 32bit OS with single core CPUs????
No you said:
Quote:
July 21 2007 all running PC's in the world...or just in the US...will be on 32bit hardware and OS's...running 32bit applications.
All PCs.
Yet again, you haven't read what I've said about running more on multicore, not necessarily multi core applications. Doing more than one thing at once. Enabling a smoother computing experience.
I feel there is a need for me to clarify what I said earlier. While I admit, i'm not the most knowledgable person in the world.. I'm still growing in that aspect every day (and I love learning new things) I still stand that as time progresses programmers can take advantage of multi-threaded environments and tailor their software to those ends.
I understand completely what is meant in terms of the inability to turn a pure singlethreaded operation into a multi-threaded one... however, if all computer applications and tasks were as simple as (3 x 7) + 4 I dont think we would have had a necessity to progress much past where we were 8 years ago.
Now, instead.. consider this:
(3 x 7) + 4 / (3 + (7 - 3)) + 24
Obviously, you cant add 4 before you multiply 3 and 7. Obviously you cant divide 25 by the result of the bottom before you execute each step accordingly. However, why cant you calculate (3 x 7) at the same time you calculate (7 - 3)? Why cant you calculate 21 + 4 at the same time you calculate 3 + 4 + 24?
Sure your example works in such a simple problem, but when you add to the complexity the potential for paralellism grows. Instead of having to calculate....
3 x 7 = 21
21 + 4 = 25
7 - 3 = 4
3 + 4 + 24 = 31
And then calculating the final result, you could cut that time down by executing several of those sub-layers (for lack of better word.. I'm certainly not a mathematician) in paralell in preperation for the final calculation.
Obviously, this is all relative. Todays computers operating at todays frequencies could execute that computation at near identicle speeds depsite being able to execute sections of it at the same time.. but for the sake of illustrating my point I believe it served its purpose.
Also, software today, in my opinion, has reached a limit on where it can grow in single threaded environments. You can only launch an application so fast before you start encroaching on the realm of prediction . Honestly, when I fire up internet explorer or firefox to browse the web... I am hard pressed to even notice it loading up. Its fast enough for me.. and I cant see the application being loaded any faster. It takes about 2 seconds on my computer to load up and begin displaying my home page. Theres literally no benefit from it loading up any faster than that.
However, being able to load 3 of those browsers at the same time now shows an increase in time. In this case 3x 2ghz cpus CERTAINLY are more powerful than 1x 6ghz cpu. If you have a base understanding for memory and memory timings you will instantly understand this. Lower latency in many situations can create more of a performance increase than increased opperating frequency. I view multi-threaded apps / operating environments (along with multi-core cpus) a great opportunity to drop latency. Just like 400mhz memory running at 3 3 3 12 can be faster than 533mhz memory running at 5 5 5 15..... 2 cpus running at 2ghz can be faster than 1cpu running at 4ghz.
Another illustration of this principle is Intels old netburst architecture with its 20 stage, and 31 stage pipes as opposed to AMD's 13 stage pipe. Intel needed higher frequencies to run at even a remotely comparable performance. However, AMD still pulled the lead due to LATENCY. Being able to execute aspects of an application in paralell lowers these "latencies".. so while the reality isnt that you are chugging through data any faster... the real world result is a quicker, smoother and more streamlined computing experience.
There in lies the potential for mult-core cpus, and multi-threaded apps. You may say there is no future, no real world benefit.. I say there is.. we only need to re-structure a large portion of how we think, work, and execute in order to accomodate it. These things take time.
I stand firm, that as our perspective as an industry in regards to multi-core, multi-cpu and multi-threaded environment continues to expand, we will find some incredible innovations arise in the way our software is programmed to take advantage of these environments. Instead of seeing how fast we can chug through a line of code, we will work out the most efficient way to process that line of code in order to make a 12 second computation into an 8 second computation into a 4 second computation. And while yes, there are a lot of examples in where you simply CANT jump the gun and proceed until you have the result of a previous calculation/computation... I believe there are many places where we can streamline the process to reduce latency and increase percieved performance. There is a lot of room to growth.
If you read all this, thanks for doing so :) and again.. i admit I'm not an engineer, scientist or any sort of math wiz, but my logic and perspective makes sense to me. Though I am always willing to stand corrected as learning is never a bad thing.
p.s. just want to add that.. while I'm not saying there are huge performance increases right now.. there may be in the future. I'll buy quad core just becaue I can (provided its not $300 more expensive or something crazy like that. Based on the intro prices of dual core in relation to single core.. I've got a feeling the price point will be respectable) and because it opens up possibilities. Programmers wont program useless code.. in this industry its usually the hardware that leads and the software that follows. After all, you dont try to move a 200 lb rock until you discover you have a cart to help you move it... and if you dont have a cart, and really want to move it.. thats when you invent the cart!
I believe thats whats happening right now with the multi thread / multi core situation. I firmly believe it will pick up as the adoption rate increases.. and as a result.. if the price isnt a huge hit (ie $100 or so for a quad core running at similar frequency to the dual core) I will buy it simply for the potential and because its my way of helping technology progress (furthering the installed base).
I knew my math analogy was gonna come back and bite me in the kiester ;) Which I why I said it was a terrible analogy (although fairly simple to grasp). If all computation were simple BEDMAS math calculations, then we'd be set.
I'm not saying there is no future, there most certainly is. It's just not as rosey as some would have us believe.
Everything you've said is largely correct. Where we differ is by gauging "how big" that improvement is going to be.
I'd wager (and this is just my personal opinion on this one), that, eventually things should balance to an overal average (across all common computing tasks) to a %30 increase in performance due to multicore. We will achieve about %30 of the theorhetical %100 performance boost. Some will obviously be higher, and others will be much lower.
%30 is nothing to sneeze at, and it comes fairly cheaply (2x2Ghz is much cheaper than 1x4Ghz). But it is not the %100 potential that many are under the impression.
There are two parts to developing multi-threaded (concurrent) applications
- conceptualizing the concurrent design. Figure out *what* will be running at the same time, where, and how. This is the hardest part. There is no coding involved. It's pure algorithms. Pure problem solving. It's impossible, (or the benefits are trivial) in many, many applications.)
-Implementation. This involves actually creating, maintaining and syncronizing the threads. This is very tricky, hard to debug/troubeshoot, and a general pain. But it's definitely do-able. This is hard, but it's the easy part, compared to the above.
To put it a different way: There are many problems where a single (linear) solution, is the best/only algorithm(solution). No amount of experience working with multi-threading application development can change that.
To use another poor analogy (heh), it's like making a highway and setting the speedlimit, to 2xC (two times the speed of light). Thats very very fast. The problem is, no car can ever go that fast. It runs up against fundamental limits to our realities. In many situations, multi-core is just like that. Something that can't be taken advantage of. (PS. This is a terribly analogy heh, so don't go bringing up Warp-Drive and all that :p)
So yeah, multi-core is great, it has it's advantages. But it's not the holy grail, and there are some things that just can't take any real advantage of it, which has nothing todo at all with the skill/experience of the software development. For linearly dependent tasks, faster single core is the only way to improve performance.
Yet again, you haven't read what I've said about running more on multicore, not necessarily multi core applications. Doing more than one thing at once. Enabling a smoother computing experience.
Sorry in one year there will still be more single core CPU-PC's (not servers) in use than multi-core CPU-PC's....no doubt about it. Heck they will probably still be selling some single core CPU's a year from now....low end stuff.
I would even wager that currently PC's with multi-core CPU's is less than 1%. Sure Intel and AMD have sold a lot of Pentium D, Core Duo and X2 CPU's in 2005-2006 (and single core CPU's) but they sold millions of single core only CPU's from say 2002-2005 a lot of which are still in use today.
It would be like saying all new cars are going to be electric from here on out. Fine....but it would take 10 years before electric cars reached 50% of what is on the road. Yes PC's last about half as long or less as cars for Joe user....so it will take half as long or less before multi core CPU's constitute 50% of what is in use.
My company has approximately 64,000 workstations/laptops. I sit on the hardware committee that approves new hardware for use. We try very hard to minimize the amount of hardware we use. We buy from HP and Dell moving to Dell mainly for the desktop. As of today we dont buy any multi-core CPUs. We will be forced very soon I am sure by Intel (we dont use AMD). That said this is where Intel is going to dump all of its netburst supply so we will probably be able to get desktops with P4s for another year and discounted prices. Fine by us we rotate are workstations every three years. Anyhow my point is the corporate world probably wont be all dual core for about 3-4 years because most companies rotate in a cycle and try to stretch things out.
To use another poor analogy (heh), it's like making a highway and setting the speedlimit, to 2xC (two times the speed of light). Thats very very fast. The problem is, no car can ever go that fast. It runs up against fundamental limits to our realities. In many situations, multi-core is just like that. Something that can't be taken advantage of. (PS. This is a terribly analogy heh, so don't go bringing up Warp-Drive and all that :p)
Aggies
Haha, dont worry.. My star-trek nerd genes stopped developing at the age of 11 so I never really got into warp-drive theory :)
Your points are taken and I understand better what you were saying. Thanks :) It'll be exciting to see where things go once we get over multi-core. I heard that quantum computing might be possible in the next 15 or so years.. though.. I wont get into that other than marvel at what it may mean.. as I know absolutely NOTHING about it :D
I agree with you in that someday quad core will be used/needed. I think in 3-4 years the majority of new software will ship in 64bit and be able to utilize multi-core CPU's...where its feasable in the application. That could be 2core...4core...or more. The follow on to Vista and Longhorn will be really multicore aware and use it alot more. Just as Vista goes a step further so will the next version in the realm of multi-core support.
Going back to my orginal point which is basically........."if your looking to upgrade or get a new PC dont get a conroe hold out for a quadcore" is not good advice as. I say this simply becuase quad core gives you nothing for some time to come over dual core.
Originally Posted by SGT Lindy There is no way I would recomend a 4 core to anyone unless.....it was cheaper than a dual core and used the same power. These days for me and most of the people I know its all about quiet PC's....quiet, cool and power friendly. I am going to get the low end 4meg conroe to replace my Pentium D so I can turn my Zalman all the way down to the slowest....quietest speed.
I totally understand what you are saying and if I was building a system for one of my parents on a budget, I'd probably just buy a Pentium D 805 for them (considering how cheap they are - they tend to keep a computer for years too, FWIW). However, they're not heavy computer users and they're certainly not bit-tech readers. I wrote this column on bit-tech because our readers spend a hell of a lot of money on computer components every year. I forget the exact amounts, but I seem to remember that the 95th %ile spends something in the region of £300 a year on graphics cards alone.
I wouldn't want one of my readers to go and buy a Core 2 Extreme X6800 now (even though it is silly fast) if there is something potentially much faster coming very soon. Obviously, it is up to readers to decide on their exact needs, but anyone who is a heavy multi-tasker - like I am - looking to spend that kind of money on a processor should hold tight for a while.
Cost-concious readers (as I stated) should probably carry on with original plans - Core 2 Duo is incredibly fast and I see little reason to consider an Athlon 64 if you are making the move to DDR2. If you're on socket 939, there are a number of choices you can make - the socket is dead though, so upgrade opportunities are going to be limited and FX-60 is the fastest processor you'll be able to buy. Although Intel is going to move forwards with quad-core processors, it's obviously going to start at the top and work its way down eventually - the push down through the rest of Intel's product line will happen during 2007.
I have only briefly skimmed through this thread so I may be covering something already mentionde, in which case I apologise...
Quote:
Originally Posted by Asphix ...stuff about maths + multi-threading...
The thing is with the example you gave, sure you could technically split that problem into different parts and work on them concurrently, I bet that it would be quicker to do it sequentially on one processor. The reason for this is that splitting the data up and then collecting it back together would be a huge communication burden (relative to the calculation). If we extrapolate this example out to applications and real world problems, sure there are some problems which do contain bits like this where things could be worked on in parallel, however it is not going to have any performance benefit due to the deficit created in the comms.
There is a fine balance between a given problem and how much the work can be divided out. Almost every application is going to have an optimum number of concurrent processes before adding any extra starts to degrade the performance.
alot of people commented on "oh its nice having quadcore, but we're not gonna use it for most apps and definitely not for gaming anyway... its overkill..."
i just have to say, quadcores is overkill NOW, yes, but by January 2007 when Vista kicks in and games such as Crysis and SupremeCommander and whatnot is out.... people will be all over this thing like peanuts and it WILL have 40-50% improvement on MOST applications (including games)
...mark my word
ps: no i didnt read ALL comments and im sure i missed a bunch of oppinions, but i did saw randomly some on each pages on this thread as well as other forums... i do not wish to *attack* or flame anybody, just putting my 2c here....
thank you...
edit: ooh i forgot one thing, i have Venice3200+ running on 2.75Ghz and wit 2Gb of RAMs, it runs Vista BETA 2 smoothly! now tat AMD cuts the price way down, im grabbing 3800+ X2 for $189CAD and wit a little overclocking, i think i can hold on to it until...i dunno... summer 2007 where kentfield is really really out in the market with decent price... so i can spend the bigbucks on some stable chipset mobo and DDR2...
so please dont say s939 is dead ;) (ok, it IS dead, but i have 7900GT and i pay games n what not and its still good until next year for most mid-level computer enthusiasts ;) )
Hello, I'm new to the forum. I was linked here by a "hot deals" forum, or whatever you would call it. People were discussing upgrading with the recent AMD price drops. It seems like a lot of people here are very well-informed on the subject, so I have a few questions to ask, if you guys don't mind. I'm sorry if they seem like newbie questions, but that's probably because by your standards, I am one.
I was wondering if now is actually a good time to go from something like a Barton 2500+ to a 3800+ dual core. I see from reading the thread there is going to be yet another processor coming out, so I'm wondering just how far the 3800+ X2 will drop, and when is going to be the sweet spot to pull the trigger on one of these things. From what I have read, currently, the 3800+ X2 is the best bang for the buck at around $155USD. Yes, I am what you would call a penny pincher. Currently, I'm just looking to upgrade my motherboard and processor, so I'm going with the ASRock board that supports AGP. Also, do any of you think it's worth it for $30USD extra to get the 4200+ X2 instead? I've been looking at benchmarks and the like, but that's just benchmarks, is there really any big difference between the 3800 and 4200 to justify the extra 30 bucks, if you're on a budget? I use my PC for some gaming(I play WoW a LOT, I'm more of an RPG/RTS player than FPS), web surfing, listening to music, AIM. I guess I'm Joe Average. ;) Alright, I think I've probably bothered you guys enough for now.
I don't know if it would help you guys to have my current system specs, but I will post them regardless, just incase. And yeah, I know my system specs suck, but it actually runs pretty decently, believe it or not.
AthlonXP Barton 2500+
ASUS A7N8X Deluxe
512x2 Corsair XMS PRO 2-2-2-5 CAS2
120GB WD 7200RPM
250GB Seagate 7200RPM
ATI Radeon 9800PRO 128MB
Antec 430w TruePower
Your specs are better than mine so I take that as an insult ;)
The x2 3800 is unlikely to drop from that price (which includes the recent price drop/slashing) for a fair while so not much point in waiting. As for the new processor you mentioned I assume you mean the core 2 duo and later the kentsfield quad version of the core 2 duo in which case that would involve getting new ram so not an option given your post (penny pincher). Is the 4200 worth $30 more, well that depends on whether you're going for the best bang for the buck or a specific level of performance. If bang for the buck is most important as it appears to be from your post then the 3800 would be a better option. The reason is that they are both from the same cpu family and have the same amount of cache which essentially means that the performance is based solely on the clock speed so working out the bang for buck isn't that hard. For example (this isn't totally accurate but is a good guideline) the 4200 is 2.2ghz which is 10% faster than the 3800 at 2ghz but costs about 19% more (185/155). Btw am using the prices you gave.
Hope that helps. Any questions just ask :)
P.S. I'm currently running an xp 2600, 512MB ram and an MX440 gpu (I don't play games so this is fine for me) and find it works ok for basic computer use like the things you mentioned but I think you'd be very pleased going to the x2 3800 and having things run both faster and smoother. Personally I'm going to wait a while longer before taking the plunge.
I appreciate the quick response, Renoir. I've decided on getting the Pentium D 840 3.2GHz, since it's dual core as well, has a faster clock speed, and I was able to grab one for 100 bucks, new, OEM. Not a bad deal, imo. Now that I'm set on my processor, I'm looking at motherboards. I'm leaning towards getting the ASUS P5VDC-MX. Seems to have pretty good reviews at newegg, and the price is right.
How much slower? Is it worth 50 more bucks in your opinion? 100 bucks seemed too cheap to pass up, so I didn't read up too much on the processor before hand. In retrospect, probably a stupid move. I had only really researched the AMD line, since they always seem to be the better 'deal'. :(
RyGuy, I'm with Tim on this one the x2 3800 would have been a better move especially considering how much more power the 840 consumes than the x2 3800. IMHO impulse buys aren't a good idea in this industry
Comments 51 to 75 of 99
You've already lost that one, im afraid.
Ok, you're points are fair enough but you've got to remember the HL2 specs are flawed since people got HL2 free with 9600s. And if there isnt continual evolvement in the industry you cant drag people kicking and screaming into the future.
HL2 isnt the most graphically demanding game on the planet and PC games take longer to develop than ever now because of the requirements of coding.
How many people vowed never to leave 98 for XP and are now on XP? It will be the same with Vista and Vista needs a serious hardware set. It will force people to upgrade. With the advent of future technology to enable people to do more, simultainiously, you will need hardware to do that. It's not so much ever about running one thing at one time, it's about having the avaliable overhead to do MANY different things and still not experience a slow down.
I lost that one??????? One year from now you think more people will be running dual core applications on a dual core CPU, on a 64bit OS than are on a 32bit OS with single core CPUs???? That is a lot of serious upgrading for a lot of people .just not going to happen in that short of a time. 3 years from now I will be it will be 50/50. There are some people and companies just giving up Windows 98 now .with lots of companies still on Windows 2000.
Most companies wont move to Vista until at least SP1. 90% of home users wont move until they need a new PC. Early vista sales will come from large corporations moving to vista and places like Best Buy and Dell selling new PC's to normal users who by a large margin make up most of the market. Hard core, overclocking bleeding edge people make up less than >1% of the market.
My company is testing Vista right now, my main workstation is running it. We are getting RC1 in August and MS has told us the Enterprise version of Vista will be ready at the end of October...we will see if they are right.
That said we are going to deploy the 32 bit version on any PC in our environment that has 512megs of ram min, with 1gig being the amount we will purchase new computers with. Our full push to approximately 64,000 users will take about 2 years. Current plans are not to deploy the 64 bit version as almost none of our applications will be 64bit save Office 2007. Right now some of their own business apps wont run on Vista like Virtual Server and Office Communicator and we are running a build past Beta 2. It will be enough for them to get the 32bit versions of these apps working let alone create full blown 64bit versions by November 1.
I run Windows XP 64bit at home and its very solid...but honestly there is nothing I need the 64bit version for. My only 64bit apps are Symantec AV 10.0 corp edition, Microsoft Defender, and Virtual Server 2005 R2. The rest is all 32 bit....single core stuff. Some drivers were a problem (camera) in the begining but most of my hardware is pretty mainstream.
At work we are moving to 64 version of Windows 2003 pretty fast. The main reason is not 64bit apps but the removal of the 4gig of virtual memory limitation. It has helped alot on some of our large file server clusters and with our SMS SQL server. However some applications like Exchange 2003 wont run on it.....but the next version Exchange 2007 will only run on the 64bit version.
It will be very interesting to see which version of Vista 64bit or 32 bit Dell and HP will ship with their mass market PC's. I would think the 64 bit versions since all new PC's will be 64bit by then....but if they are worried about compatibility maybe not????
In a properly multithreaded application, 2x 2GHz cores will perform almost identically to 1x 4GHz. When I moved from a single- to a dual-core system, of about the same clock speeds, my speed in threaded applications was almost exactly doubled.
As to the various 64-bit comments... are you running 4GB or more of RAM? If not, you have very little to gain by moving to a 64-bit OS. 2^32 just so happens to be the number of bytes in 4GB, and it's no coincidence. You can be pretty certain that the limit of addressable memory with a 64-bit system is 16 EB - 16 billion gigabytes (well, disregarding the 1024 vs 1000 thing). I don't think it would be an unreasonable estimate to say that amount is probably on par with all of the RAM on the planet, and almost certainly within the country.
Obviously there are some skeptical users on here. Of those who are - have you USED a SMP machine, be it multicore or multiprocessor? If yes, then I'm sure you have your reasons for thinking what you do. My dual-core machine was my first SMP experience. And quite literally after first boot, I immediately decided that I'd never go back. Everything is smoother in general usage, and CPU-intensive tasks never leave the machine in an unusable state. Heck, I've gamed with a recode running in the background, and barely noticed - sure, my FPS was a tad lower than usual, but it certainly was no slideshow. The only way I can get my computer in a somewhat painful state to use is when I've got Xvid transcoding going on while defragging and ripping apart another DVD. I've had occasions where I'd have loved a quad-core system, but the only times I felt such a way was when I was cranking on my machine like it's never seen before.
So, long story short, try it before you knock it.
I agree with Tim regarding everything you said. Multi-core is certainly not a dead end. There are definite, tangible and immediate benefits to going dual core. Both for specifc applications (FWIW I'd consider "media compression" ~= "media encoding" :p), and also for general computing.
But remember the main topic of the thread, "Why buy 2cores, when 4 cores are comming out in 6months?"
Let me put out the arguments that I believe leads to the conclusion that *most* people (who read bit-tech, not just most people in general) shouldn't wait.
- Not all apps can see benefits to multithreading. Those that do, the performance increases are often not always %100. "The benefits of dual core aren't %100"
- For those that do see benefits to multithreading, those benefits will become even smaller when going from 2->4 cores (than when going from 1->2 cores). "The benefits of quad-core are even less of an improvement".
With the above two points, I'd say it's not worth waiting. That quad core doesn't add *enough* to the table to justify waiting. Obviously there is improvement, it's just not enough to warrant delaying an upgrade. Especially when compared to the jump that is/was Conroe.
Thats my whole point. But there is a definite trend in current opinions that extole the virtues of multi-core/paralellization. It's treated as if 2cores essentially result in one "twice-as-fast" single core. And my contention is that is quite far from the truth. Multi-core has it's benefits, but in specific circumstances which aren't usually in the majority.
Again, if you were to give me a 4core chip, I'd take it without hesitation. It can't hurt, and there are benefits to be realized. But it's not the end all and be all, and more importantly the gains from 2->4 are less than from 1->2
Otherwise we'd all be running massive parallel arrays of p166's in our homes :p
Aggies
I fully agree with you. I can tell when I am doing alot of work on box that is single core....like my laptop with a Pentium M...compared to my work desktop box that has hyperthreading. Bouncing back and forth between 5-7 applications cutting and pasting there are less pauses with the Hyperthreaded CPU.....and even slightly better with my Pentium D at home.
That said I cant run a Virus Scan and play BF2 at the same time with out serious performance hits in both...even with my Pentium D 930 because niether of those applications are dual core aware.....XP will try to spread the threads across both cores but it does not work very well.
If I were playing just BF2 and not doing anything else....having a single core or dual core would not matter.....and that is true for 95% of applications out there.
If I were building a new PC for a friend today...I would go for the lowest Conroe that has 4meg of cache, E6400?? Because I do know that dual core gives you some advantages...and just makes things smoother overall. If this person was my mother-n-law or someone that just surfs the web, uses email, downloads pictures and uses Quicken....then I would go for the cheapest conroe they will make...even a single core celeron version of they are going to make/label one. That type of person is probably moving off of some Pentium III or P4 Celeron anyway and even the slowest conroe would be a huge boost.
There is no way I would recomend a 4 core to anyone unless.....it was cheaper than a dual core and used the same power. These days for me and most of the people I know its all about quiet PC's....quiet, cool and power friendly. I am going to get the low end 4meg conroe to replace my Pentium D so I can turn my Zalman all the way down to the slowest....quietest speed.
No you said:
All PCs.
Yet again, you haven't read what I've said about running more on multicore, not necessarily multi core applications. Doing more than one thing at once. Enabling a smoother computing experience.
I understand completely what is meant in terms of the inability to turn a pure singlethreaded operation into a multi-threaded one... however, if all computer applications and tasks were as simple as (3 x 7) + 4 I dont think we would have had a necessity to progress much past where we were 8 years ago.
Now, instead.. consider this:
(3 x 7) + 4 / (3 + (7 - 3)) + 24
Obviously, you cant add 4 before you multiply 3 and 7. Obviously you cant divide 25 by the result of the bottom before you execute each step accordingly. However, why cant you calculate (3 x 7) at the same time you calculate (7 - 3)? Why cant you calculate 21 + 4 at the same time you calculate 3 + 4 + 24?
Sure your example works in such a simple problem, but when you add to the complexity the potential for paralellism grows. Instead of having to calculate....
3 x 7 = 21
21 + 4 = 25
7 - 3 = 4
3 + 4 + 24 = 31
And then calculating the final result, you could cut that time down by executing several of those sub-layers (for lack of better word.. I'm certainly not a mathematician) in paralell in preperation for the final calculation.
Obviously, this is all relative. Todays computers operating at todays frequencies could execute that computation at near identicle speeds depsite being able to execute sections of it at the same time.. but for the sake of illustrating my point I believe it served its purpose.
Also, software today, in my opinion, has reached a limit on where it can grow in single threaded environments. You can only launch an application so fast before you start encroaching on the realm of prediction . Honestly, when I fire up internet explorer or firefox to browse the web... I am hard pressed to even notice it loading up. Its fast enough for me.. and I cant see the application being loaded any faster. It takes about 2 seconds on my computer to load up and begin displaying my home page. Theres literally no benefit from it loading up any faster than that.
However, being able to load 3 of those browsers at the same time now shows an increase in time. In this case 3x 2ghz cpus CERTAINLY are more powerful than 1x 6ghz cpu. If you have a base understanding for memory and memory timings you will instantly understand this. Lower latency in many situations can create more of a performance increase than increased opperating frequency. I view multi-threaded apps / operating environments (along with multi-core cpus) a great opportunity to drop latency. Just like 400mhz memory running at 3 3 3 12 can be faster than 533mhz memory running at 5 5 5 15..... 2 cpus running at 2ghz can be faster than 1cpu running at 4ghz.
Another illustration of this principle is Intels old netburst architecture with its 20 stage, and 31 stage pipes as opposed to AMD's 13 stage pipe. Intel needed higher frequencies to run at even a remotely comparable performance. However, AMD still pulled the lead due to LATENCY. Being able to execute aspects of an application in paralell lowers these "latencies".. so while the reality isnt that you are chugging through data any faster... the real world result is a quicker, smoother and more streamlined computing experience.
There in lies the potential for mult-core cpus, and multi-threaded apps. You may say there is no future, no real world benefit.. I say there is.. we only need to re-structure a large portion of how we think, work, and execute in order to accomodate it. These things take time.
I stand firm, that as our perspective as an industry in regards to multi-core, multi-cpu and multi-threaded environment continues to expand, we will find some incredible innovations arise in the way our software is programmed to take advantage of these environments. Instead of seeing how fast we can chug through a line of code, we will work out the most efficient way to process that line of code in order to make a 12 second computation into an 8 second computation into a 4 second computation. And while yes, there are a lot of examples in where you simply CANT jump the gun and proceed until you have the result of a previous calculation/computation... I believe there are many places where we can streamline the process to reduce latency and increase percieved performance. There is a lot of room to growth.
If you read all this, thanks for doing so :) and again.. i admit I'm not an engineer, scientist or any sort of math wiz, but my logic and perspective makes sense to me. Though I am always willing to stand corrected as learning is never a bad thing.
p.s. just want to add that.. while I'm not saying there are huge performance increases right now.. there may be in the future. I'll buy quad core just becaue I can (provided its not $300 more expensive or something crazy like that. Based on the intro prices of dual core in relation to single core.. I've got a feeling the price point will be respectable) and because it opens up possibilities. Programmers wont program useless code.. in this industry its usually the hardware that leads and the software that follows. After all, you dont try to move a 200 lb rock until you discover you have a cart to help you move it... and if you dont have a cart, and really want to move it.. thats when you invent the cart!
I believe thats whats happening right now with the multi thread / multi core situation. I firmly believe it will pick up as the adoption rate increases.. and as a result.. if the price isnt a huge hit (ie $100 or so for a quad core running at similar frequency to the dual core) I will buy it simply for the potential and because its my way of helping technology progress (furthering the installed base).
I knew my math analogy was gonna come back and bite me in the kiester ;) Which I why I said it was a terrible analogy (although fairly simple to grasp). If all computation were simple BEDMAS math calculations, then we'd be set.
I'm not saying there is no future, there most certainly is. It's just not as rosey as some would have us believe.
Everything you've said is largely correct. Where we differ is by gauging "how big" that improvement is going to be.
I'd wager (and this is just my personal opinion on this one), that, eventually things should balance to an overal average (across all common computing tasks) to a %30 increase in performance due to multicore. We will achieve about %30 of the theorhetical %100 performance boost. Some will obviously be higher, and others will be much lower.
%30 is nothing to sneeze at, and it comes fairly cheaply (2x2Ghz is much cheaper than 1x4Ghz). But it is not the %100 potential that many are under the impression.
There are two parts to developing multi-threaded (concurrent) applications
- conceptualizing the concurrent design. Figure out *what* will be running at the same time, where, and how. This is the hardest part. There is no coding involved. It's pure algorithms. Pure problem solving. It's impossible, (or the benefits are trivial) in many, many applications.)
-Implementation. This involves actually creating, maintaining and syncronizing the threads. This is very tricky, hard to debug/troubeshoot, and a general pain. But it's definitely do-able. This is hard, but it's the easy part, compared to the above.
To put it a different way: There are many problems where a single (linear) solution, is the best/only algorithm(solution). No amount of experience working with multi-threading application development can change that.
To use another poor analogy (heh), it's like making a highway and setting the speedlimit, to 2xC (two times the speed of light). Thats very very fast. The problem is, no car can ever go that fast. It runs up against fundamental limits to our realities. In many situations, multi-core is just like that. Something that can't be taken advantage of. (PS. This is a terribly analogy heh, so don't go bringing up Warp-Drive and all that :p)
So yeah, multi-core is great, it has it's advantages. But it's not the holy grail, and there are some things that just can't take any real advantage of it, which has nothing todo at all with the skill/experience of the software development. For linearly dependent tasks, faster single core is the only way to improve performance.
Aggies
Sorry in one year there will still be more single core CPU-PC's (not servers) in use than multi-core CPU-PC's....no doubt about it. Heck they will probably still be selling some single core CPU's a year from now....low end stuff.
I would even wager that currently PC's with multi-core CPU's is less than 1%. Sure Intel and AMD have sold a lot of Pentium D, Core Duo and X2 CPU's in 2005-2006 (and single core CPU's) but they sold millions of single core only CPU's from say 2002-2005 a lot of which are still in use today.
It would be like saying all new cars are going to be electric from here on out. Fine....but it would take 10 years before electric cars reached 50% of what is on the road. Yes PC's last about half as long or less as cars for Joe user....so it will take half as long or less before multi core CPU's constitute 50% of what is in use.
My company has approximately 64,000 workstations/laptops. I sit on the hardware committee that approves new hardware for use. We try very hard to minimize the amount of hardware we use. We buy from HP and Dell moving to Dell mainly for the desktop. As of today we dont buy any multi-core CPUs. We will be forced very soon I am sure by Intel (we dont use AMD). That said this is where Intel is going to dump all of its netburst supply so we will probably be able to get desktops with P4s for another year and discounted prices. Fine by us we rotate are workstations every three years. Anyhow my point is the corporate world probably wont be all dual core for about 3-4 years because most companies rotate in a cycle and try to stretch things out.
Haha, dont worry.. My star-trek nerd genes stopped developing at the age of 11 so I never really got into warp-drive theory :)
Your points are taken and I understand better what you were saying. Thanks :) It'll be exciting to see where things go once we get over multi-core. I heard that quantum computing might be possible in the next 15 or so years.. though.. I wont get into that other than marvel at what it may mean.. as I know absolutely NOTHING about it :D
I agree with you in that someday quad core will be used/needed. I think in 3-4 years the majority of new software will ship in 64bit and be able to utilize multi-core CPU's...where its feasable in the application. That could be 2core...4core...or more. The follow on to Vista and Longhorn will be really multicore aware and use it alot more. Just as Vista goes a step further so will the next version in the realm of multi-core support.
Going back to my orginal point which is basically........."if your looking to upgrade or get a new PC dont get a conroe hold out for a quadcore" is not good advice as. I say this simply becuase quad core gives you nothing for some time to come over dual core.
I wouldn't want one of my readers to go and buy a Core 2 Extreme X6800 now (even though it is silly fast) if there is something potentially much faster coming very soon. Obviously, it is up to readers to decide on their exact needs, but anyone who is a heavy multi-tasker - like I am - looking to spend that kind of money on a processor should hold tight for a while.
Cost-concious readers (as I stated) should probably carry on with original plans - Core 2 Duo is incredibly fast and I see little reason to consider an Athlon 64 if you are making the move to DDR2. If you're on socket 939, there are a number of choices you can make - the socket is dead though, so upgrade opportunities are going to be limited and FX-60 is the fastest processor you'll be able to buy. Although Intel is going to move forwards with quad-core processors, it's obviously going to start at the top and work its way down eventually - the push down through the rest of Intel's product line will happen during 2007.
There is a fine balance between a given problem and how much the work can be divided out. Almost every application is going to have an optimum number of concurrent processes before adding any extra starts to degrade the performance.
well crap, if this dude ( http://www.xtremesystems.org/forums/showthread.php?t=107092 ) gets 6,000+ CPU score on 3dmark06, its definitely gonna do something good for Q1-2007 games like Crysis, would it not?
i just have to say, quadcores is overkill NOW, yes, but by January 2007 when Vista kicks in and games such as Crysis and SupremeCommander and whatnot is out.... people will be all over this thing like peanuts and it WILL have 40-50% improvement on MOST applications (including games)
...mark my word
ps: no i didnt read ALL comments and im sure i missed a bunch of oppinions, but i did saw randomly some on each pages on this thread as well as other forums... i do not wish to *attack* or flame anybody, just putting my 2c here....
thank you...
edit: ooh i forgot one thing, i have Venice3200+ running on 2.75Ghz and wit 2Gb of RAMs, it runs Vista BETA 2 smoothly! now tat AMD cuts the price way down, im grabbing 3800+ X2 for $189CAD and wit a little overclocking, i think i can hold on to it until...i dunno... summer 2007 where kentfield is really really out in the market with decent price... so i can spend the bigbucks on some stable chipset mobo and DDR2...
so please dont say s939 is dead ;) (ok, it IS dead, but i have 7900GT and i pay games n what not and its still good until next year for most mid-level computer enthusiasts ;) )
Who on earth has a machine that only has one application or system service running at a time?!
I was wondering if now is actually a good time to go from something like a Barton 2500+ to a 3800+ dual core. I see from reading the thread there is going to be yet another processor coming out, so I'm wondering just how far the 3800+ X2 will drop, and when is going to be the sweet spot to pull the trigger on one of these things. From what I have read, currently, the 3800+ X2 is the best bang for the buck at around $155USD. Yes, I am what you would call a penny pincher. Currently, I'm just looking to upgrade my motherboard and processor, so I'm going with the ASRock board that supports AGP. Also, do any of you think it's worth it for $30USD extra to get the 4200+ X2 instead? I've been looking at benchmarks and the like, but that's just benchmarks, is there really any big difference between the 3800 and 4200 to justify the extra 30 bucks, if you're on a budget? I use my PC for some gaming(I play WoW a LOT, I'm more of an RPG/RTS player than FPS), web surfing, listening to music, AIM. I guess I'm Joe Average. ;) Alright, I think I've probably bothered you guys enough for now.
I don't know if it would help you guys to have my current system specs, but I will post them regardless, just incase. And yeah, I know my system specs suck, but it actually runs pretty decently, believe it or not.
AthlonXP Barton 2500+
ASUS A7N8X Deluxe
512x2 Corsair XMS PRO 2-2-2-5 CAS2
120GB WD 7200RPM
250GB Seagate 7200RPM
ATI Radeon 9800PRO 128MB
Antec 430w TruePower
The x2 3800 is unlikely to drop from that price (which includes the recent price drop/slashing) for a fair while so not much point in waiting. As for the new processor you mentioned I assume you mean the core 2 duo and later the kentsfield quad version of the core 2 duo in which case that would involve getting new ram so not an option given your post (penny pincher). Is the 4200 worth $30 more, well that depends on whether you're going for the best bang for the buck or a specific level of performance. If bang for the buck is most important as it appears to be from your post then the 3800 would be a better option. The reason is that they are both from the same cpu family and have the same amount of cache which essentially means that the performance is based solely on the clock speed so working out the bang for buck isn't that hard. For example (this isn't totally accurate but is a good guideline) the 4200 is 2.2ghz which is 10% faster than the 3800 at 2ghz but costs about 19% more (185/155). Btw am using the prices you gave.
Hope that helps. Any questions just ask :)
P.S. I'm currently running an xp 2600, 512MB ram and an MX440 gpu (I don't play games so this is fine for me) and find it works ok for basic computer use like the things you mentioned but I think you'd be very pleased going to the x2 3800 and having things run both faster and smoother. Personally I'm going to wait a while longer before taking the plunge.