Windows Vista SP1 Core Performance

Did you guys not disable Readyboost when running boot time benchmarks? Readyboot - the boot optimisation service is inside of readyboost.

Originally Posted by cjmUK
I'd rather you spend some extra time running benchmarks that *may* be somewhat comparable than spending less time running benchmarks that cannot be comparable ever.

Readyboost is a crutch for systems without much RAM. But Superfetch is an integral part of Vista - disabling it is like tieing one of Vista's arms behind it's back.

You could even say it's somewhat unfair that you compare both OS's with the same hardware - to be truly fair, you should compare both OS's 'recommended hardware' configurations. But OK, we can let this one slip, but not the omission of SuperFetch...

[fair play - in an SP1 vs Pre-SP1 test, one would assume each system is equally crippled by the disabling of SuperFetch]

As for the effort involved in preparing the test environment, I don't see why GHOSTing can't be employed to good effect. It only takes a couple of hours to wipe a system. I realise that you have logistical and time constraints, but you can't jeopardize the integrity of BT to save a bit of time...

OK, I'm lecturing now so I'll stop. :)

It's not jeopardising anyone's integrity - it's something that was being discussed long before we actually moved to Vista for benchmarking. It wasn't until May that we actually made the move and in the run up to that, we weren't just sitting there twiddling our thumbs. We spent a lot of time talking with the industry (AMD's, Intel's and Nvidia's benchmarking labs) about how to obtain accurate benchmarks under Vista because we had a lot of concerns with the way the OS does some things.

What you see is the result of that effort--consistency is what is paramount for us when it comes to testing hardware, otherwise you're making assumptions and conclusions about a product/architecture from the analysis you've done on the benchmarks... when the differences could actually just be a result of Vista doing its own thing. And that is when technical inaccuracies and false analysis creeps in... I'd say that is more likely to jeopardise our integrity than choosing to disable a feature in order to deliver accurate benchmarks.

WRT systems without enough memory - it's cheap as chips and we're testing everything with a minimum of 2GB... which costs all of £35 these days. Even 4GB costs just £80-90, depending on brand and whether or not you opt for faster timings than 5-5-5.

I tried Vista with 1GB and it was painful, so yes of course ReadyBoost will make a difference there.

/cry
The ONLY thing I wanted (nay, needed) was an improvement in copying medium sized (~700MB) from HDD to NAS. :(

Originally Posted by Tim S
Enable Advance Performance reintroduces an old bug from Windows 3.11 and can cause data corruption and data loss if your drive loses power. See here: http://technet.microsoft.com/en-gb/magazine/cc162475.aspx

Well, last time I checked I have lots of backups, and in here in Canada we have most of the time electricity... Maybe for you guys it cuts every 30min... not here, perhaps one of those backup batteries should be installed for your computers ;) Also both ways, you can have data loss. Also, it is not a bug, it is an issue. A bug is fixable, and issue require a re-think of how to make the feature works. If MS re-integrated that feature, it is for a good reason, perhaps they integrated some protection system, maybe not full proof, however still there. Else I could just do benchmark test and set the hard drive to removable, and I can show you guys how Windows Vista is painfully slow.
And i am sure it is safer than a RAID setup.

Originally Posted by Tim S
It's not jeopardising anyone's integrity - it's something that was being discussed long before we actually moved to Vista for benchmarking. It wasn't until May that we actually made the move and in the run up to that, we weren't just sitting there twiddling our thumbs. We spent a lot of time talking with the industry (AMD's, Intel's and Nvidia's benchmarking labs) about how to obtain accurate benchmarks under Vista because we had a lot of concerns with the way the OS does some things.

What you see is the result of that effort--consistency is what is paramount for us when it comes to testing hardware, otherwise you're making assumptions and conclusions about a product/architecture from the analysis you've done on the benchmarks... when the differences could actually just be a result of Vista doing its own thing. And that is when technical inaccuracies and false analysis creeps in... I'd say that is more likely to jeopardise our integrity than choosing to disable a feature in order to deliver accurate benchmarks.

Such results would just make Microsoft to force these technology in and turned on. What is Windows 7 forced these technology to be turned on... than what? Will you say that Win7 is superior in performance then Vista, even thus the results matched Vista with these features turned on.

Also, I don't see you guys disable Read and write cache in WinXP?! Disable those, and you WILL SEE how WinXP is soooooooo slllloooowwwwww.

What should have been done, is present BOTH results, and let the reader decide what best fit their needs. Instead of thinking for weeks how it could be done and show "false" results, you could have presented both results.

It probably is safer than RAID.... 0. But I have never once recommended anyone bothers running RAID 0 - in fact, I believe every system review we've done where there's a RAID 0 array has at the very least had the point that RAID 0 is not recommended for all but the e-peen seekers. :)

Well you should do the same... enable it, show the results, and go... "not recommend if you don't have backups of your data and battery as it could lead to data loss if you lose power. As it is suggested by Microsoft in the description bellow the options." Than the user will decide yes/no, but for now, many users don't see the benchmark of that feature, so they won't even bother trying it, and see a nice performance increase, OR people like me that thinks that their is a performance gain, but in fact it is insignificant and not worth the risk.

Let me also add, that it should also tested with laptops, as perhaps Microsoft main focus for SP1 was for laptop instead of desktop.

Is the Thecus N2100 the best way to test network transfers? Since most users will be transferring between PCs or a full blown server, this NAS system with it's slower processor and other bottlenecks could give a false impression of the usual transfer experience. Also, the release notes specifically mention improvements between two SP1 machines, which are not tested in this review.

I hope all these comments from readers don't give the impression that we're not happy to see a report from you guys, I guess it's just a touchy subject!

i whent from win98 to XP and i do prefer it, XP with SP2 was the best thing M$ ever did for XP, pre SP2 all you had to do is have an router or have an firewall turnd on and you be ok (MSblaster :) )

i used windows 2000 for my server or i did untill i could not find any drivers for it after i reloaded windows got 2003 64-bit on there now mosty works ok get long pauses when playing video on my pc when the server is moveing files around (600mb unpack snewsbin)

Originally Posted by ashchap
I hope all these comments from readers don't give the impression that we're not happy to see a report from you guys, I guess it's just a touchy subject!

We've always said that all feedback is appreciated--whether good or bad--as long as it is constructive. Things haven't changed here and feedback is being taken on board - I'd rather have an interactive process where we make sure we're delivering the right information to you, the readers, and we explain why we've done things certain ways. I've always loved bit-tech for this reason, because you guys are so vocal and I don't ever want that to change! :)

I can't answer everything right now as I've got a lot to get done in the next 72 hours before I fly to the Far East for a couple of weeks of visiting various manufacturers out there, but rest assured that nobody is being ignored.

superfetch works wunders for opera browser when you got 20-30 tabs on it when you open opera way faster then XP as opera stores the last page that was on each tab (firefox is just an cheap addon that for recovery only if it crashed and it only links to the page)

---

one thing thay Realy need to fix if i do not have my X-FI in my pc i can copy files from my server @ 125MB/s (1gb network)
thats Strate line NO dips 2xraid to 2xraid both PCs useing Nvidia based network cards so its HW cards
with the sound card in Soon as windows make any sound like an Ding speed drops way down to 20MB/s, the only fix is to reboot windows

i could just remove the card but on board sound is realy bad compeared to X-FI due to not been able to locate where the foot steps are comming from

firefox is just an cheap addon that for recovery only if it crashed and it only links to the page)

You can do that without an add-on in Firefox. Look under "Main" tab in the options.

As for the rest, this is Creative fault, not Microsoft. I suffer the same thing as you. If I uninstall my Creative drivers and card, my computer runs faster, network speed increase, and boot up faster. This was the same under XP. I still don't understand how review sites did not discover these issues... I hope that now, review sites will look for such things.
At the begging I thought it was just me, but Creative forum is bombarded with these issues.

it does not work right in firefox as i lost all my tabs in firefox as i had an lame popup box that i closed last as it only remember the last firefox window thats open in opera Opera Never allows an second copy to be launched unless you right click and tell it so

i dono about the creative and network speeds should not affect each other really (COD4/netgear is bugging me i refresh me server list and me router locks up heh)

the Vista SP1 works well as i loaded it onto 2-3 computers and one in my VMware box to test if Bios patch work around still works and does not so cant use SP1 at this time (i own an vista ultimate 64 OEM disk and key but not going tho the hassle of activating windows all day)

one thing i was hoping thay fix was superfetch it needs to be i/o priority so if an program is accessing the hard disk superfetch should stop filling the cache until the program stops messing with the hard disk this is more important for laptop users as it can bog the pc down for up to 5-8 mins if 2gb of ram is used, with systems like mine 4gb using 4x hdds in raid 0 hides the fact windows is doing stuff (laptop users should use the sleep option in vista, never seems to work on desktops BSOD :) most likey me sound card)

I too am somewhat worried at the idea of a key feature such as superfetch being turned off.

However, as I understand it, this will give a "worst-case" scenario which assumes a fresh install of an app, before superfetch has had a chance to "see" the application and develop an idea of the patterns of use for that application.

Thus, the tests being run by bit-tech would approximate the first ever run of an application on a vista computer, with successive tests yielding better results.

Actually, is there an article in there, or has it already been done (Superfetch and its effects on gaming performance)?

I haven't done any benchmarking but from my personal experience on my home network of two hugely different Vista desktops and a nasty old slow Vista laptop I'd have to disagree with the results of your tests. On all three computers I have seen vast inprovements in performance across many of those categories.

Just for example pre-SP1 bootup on my Vista box was torture, I had to sit and wait a few minutes for my network to connect. I eventually disabled all startup programs AVG, anti-spyware and even down to Windows Sidebar just to be able to use my PC immediately on boot.

After SP1 bootup is almost instant, I can turn on the PC and watch as my system tray fills with icons in a second flat. I don't even bother stopping worthless items like Adobe speed launcher because it's so fast now.

Same with file copying, I would get up and go for a walk if I knew I was going to move some files around Pre-SP1, now I can just wait while the copy completes.

Originally Posted by =cJ=
I too am somewhat worried at the idea of a key feature such as superfetch being turned off.

However, as I understand it, this will give a "worst-case" scenario which assumes a fresh install of an app, before superfetch has had a chance to "see" the application and develop an idea of the patterns of use for that application.

Thus, the tests being run by bit-tech would approximate the first ever run of an application on a vista computer, with successive tests yielding better results.

Actually, is there an article in there, or has it already been done (Superfetch and its effects on gaming performance)?

Welcome!

When I first tested with Vista I didn't turn either off and as a consequence results were massively inconsistent, sometimes taking two/three times longer, sometimes shorter, on consecutive runs of the same real world programs we use for motherboard testing. MS may have improved it since, but compared to XP it still couldn't come close and it made seeing the fine differences in testing a nightmare. If you've got more than enough memory (2GBs are dirt cheap) and a defragged, fast hard drive I'd suggest the differences are questionable. It's also dependent on how you use a PC.

Originally Posted by Bindibadgi
If you've got more than enough memory (2GBs are dirt cheap) and a defragged, fast hard drive I'd suggest the differences are questionable. It's also dependent on how you use a PC.

I'm not sure what you are suggesting here... that if you have enough memory and a defragged HDD that you don't need Superfetch?

Defragged disks are good - no argument there. But the more RAM you have, the better SuperFetch performs. Many people think that if their most demanding app needs less than 2GB, then 2GB RAM is all that is needed, but if you have, say, 4GB, you can keep 2GB clear for your apps and 2GB for the OS & SuperFetch. The more RAM yo have, the better SuperFetch can prepare for your next action(s).

As I read all these comments bashing Vista like a bad cocaine habit I can't help wondering...

I took my first tentative trials of Vista while it was still going by its codename; Longhorn. As Vista eventually matured and went RTM I got hold of and installed the UItimate version - and I have never looked towards my XP discs again. Sure, driver support was rubbish at first and it was riddled with bugs - but this was also the case with XP when it was fresh on the shelves. (I also tried out XP while it was still going under the name "Whistler").

I do not have the benchmarks to prove it, but I am certain my system runs far better on Vista than it did on XP - it most definitely feels way more responsive. I have none of the common gaming issues seeming to plague Vista users everywhere if we are to believe the hype, and on a whole it is my honest opinion that Vista is a far better system than XP ever was. I agree that SP1 seemingly did not have a profound effect on gaming performance, but booting time and general core system performance is another matter. I am quite certain (again I don't have the numbers to prove it) that it boots noticeably faster than before, and I am also certain that file management is faster. As I said; the whole system feels much more alive and responsive. I am currently running the 32bit version from a Raptor RAID0. As soon as my Reaper 4Gb kit arrives I will be going 64bit finally.

I do have one strange problem though; I am only able to use the 169.01 (I'm at work so I may not have remembered the version number correctly) version Nvidia driver for my GTS. I have tried all the newer ones, and each attempt with a newer driver presents me with a black screen after Vista has loaded. I get to see the loading screen and everything up to that point, but when it's going to desktop the screen goes black. My monitor is the Benq G2400W. Any ideas?

Originally Posted by cjmUK
I'm not sure what you are suggesting here... that if you have enough memory and a defragged HDD that you don't need Superfetch?

Defragged disks are good - no argument there. But the more RAM you have, the better SuperFetch performs. Many people think that if their most demanding app needs less than 2GB, then 2GB RAM is all that is needed, but if you have, say, 4GB, you can keep 2GB clear for your apps and 2GB for the OS & SuperFetch. The more RAM yo have, the better SuperFetch can prepare for your next action(s).

Sorry - if you have enough memory ReadyBoost is negligible iirc.

Superfetch may chuck lots into ram, but it's using the hard drive to pull everything off when it likes, which generally messes with results because it's just not clever enough. It wholly depends on what software you use and how you use it.

But the point is - for OUR RESULTS, the small differences would be lost or I would be here forever trying to get accurate, reliable data.

Excellent article overall,

I think the underlying picture that the world over is getting to is that Vista is a minor upgrade to XP and has not gone far enough in raising the users overall experience with the OS.

quote: "We can do it on Internet and FTP transfers so why not when the drive is within arm’s reach not the other side of the world? "

And this SP1 again underlines these points and in fact, the major performance increases should be seen as bugs that needed to be fixed as the poor performance pre-SP1 was really intolerable. Now it is just acceptable.

Something must change at M$ otherwise the market is ripe for a major shakeup should somebody release an OS that can cover the same user/computer base as Vista does.

Originally Posted by Bindibadgi
Sorry - if you have enough memory ReadyBoost is negligible iirc.

Yeah, we're not debating that.

Superfetch may chuck lots into ram, but it's using the hard drive to pull everything off when it likes, which generally messes with results because it's just not clever enough. It wholly depends on what software you use and how you use it.

If you have slow or poorly maintained disks, the effect of SuperFetch will be somewhat muted because the drives won't be able to adequately feed both the live applications and SuperFetch. However if you have a 7200rpm SATA II drive that is defragged, most of the time SuperFetch will be be able to pull enough stuff off the disk to feed the live application and some anticipated future applications.]

SuperFetch works. It's not perfect, but in most circumstances it is of benefit. I've yet to read anywhere that SuperFetch isn't an improvement on any previous caching tecnologies and I've yet to read of any circumstances where disabling SuperFetch would be beneficial.

But the point is - for OUR RESULTS, the small differences would be lost or I would be here forever trying to get accurate, reliable data.

The point is you have foregone potentially variable test results in favour of *inaccurate, unreliable* but consistent results. It was a decision of convenience.

I understand what you guys are saying about the difficulty of (and effort involved in) preparing a suitable test environment, but given the nature of technologies like SuperFetch there are no shortcuts.

The only thing I could have recommended was that you create basic XP and Vista test machines, and use each for similar purposes for a day, and then run the tests. Sure, depending what each machine had been used for, the impact of SuperFetch would have varied, but such is life. They would have been valid, real-world results though.

...Vista is a minor upgrade to XP

I think this sums up the problem with Vista. Too few people actually realise what has changed in Vista - Vista is a major shake-up of MS's x86/x64 platforms. There are radical changes under the bonnet that will form the foundations of future software, but because there isn't enough bling to appeal to Joe Public. XP set the bar so high that peoples expectations have become rather unrealistic. Things dont ever get faster... if there is more power at hand it simply means that we do more in the time available. If anyone remembers WordPerfect 5.1 for DOS... it ran like the proverbial off the shovel and did everything that we needed. But we are quite happy to use the bloated behemoth that is Word 2007. Things get bigger, not faster.

The fact is that Vista was in a better state at launch than XP was. It's a good OS. Yes, it bloated like all MS software, but that is the nature of the beast from Redmond. Vista also performs at least as good as XP did on the standard hardware that was available at the time of launch.

I realise that I sound like a hardcore Vista evangelist, but the fact is that there is plenty about Vista that I am critical of, but nobody has got far enough past "Meh, it's slower than XP!" to actually discuss it with me. I'm pissed off with the wasted opportunity presented by UAC, I think there are too many levels under the control panel, I think the network tools are over-simplified, yet confusing, and I disagree with them promoting Sleep in an age when we are told not to leave our TVs on Standby.

But... none of that is relevant if you cant even install SP1...

Originally Posted by cjmUK
The only thing I could have recommended was that you create basic XP and Vista test machines, and use each for similar purposes for a day, and then run the tests. Sure, depending what each machine had been used for, the impact of SuperFetch would have varied, but such is life. They would have been valid, real-world results though.

You're recommending a massive amount of testing time no serious site has spare. As it is that article took me two solid 12 hour days obtain accurate results. What you're asking for is the creation of an artificial and massive variable - how long do I use the machine, how much do I use the software, do I defrag during or after it, what about other people installing other programs, what about the processes done WITHIN those programs, do I use a machine how others do, how many programs should be installed to make it "real world"? It's an impossible number to quantify accurately.

The results are not inaccurate and unreliable, what you're asking for is infeasible, inaccurate and unreliable for the sake of a single caching feature. What would comparative results tell you if they were all over the place? Since we're comparing within the Vista environment it doesn't make any difference since, in theory, it should work the same throughout. We're not comparing Vista to OSX here.

Originally Posted by vaderag
But... none of that is relevant if you cant even install SP1...

True. :(

Originally Posted by Bindibadgi

Since we're comparing within the Vista environment it doesn't make any difference since, in theory, it should work the same throughout. We're not comparing Vista to OSX here.

To be fair, I'm thinking ahead to the upcoming XP vs Vista article. Disabling Superfetch in the last two benchmarks is clearly more forgiveable, though we can't be sure on it's impact.

Some examples of how other sites benchmarked Vista:

http://www.tomshardware.com/2007/01/29/xp-vs-vista/page2.html
http://www.extremetech.com/article2/0,1697,2096942,00.asp
http://www.pcworld.com/article/id,128305-page,1-c,vistalonghorn/article.html

No other benchmarking article mentions disabling SuperFetch, and Toms hardware even goes to the trouble of explaining how their test environment was prepared, on order to take advantage of the real-world benefits of SuperFetch.

I don't have a monopoly on wisdom, and I can't begin to predict the performance difference with SuperFetch enabled, but it seems clear that BT is the only site that disables a key feature like SuperFetch in such a situation.

Anyway, I've said enough; you're either convinced or you are not, and I'm sure everybody else is sick of me hogging this thread, so I guess I've said my bit.