bit-tech.net

AMD Opteron 6174 vs Intel Xeon X5650 Review

Comments 26 to 50 of 50

Reply
barndoor101 31st March 2010, 21:48 Quote
anandtech are good for stuff like that comdot - they did a really good set of benchmarks of these chips with virtualisation in mind.
Jasio 31st March 2010, 22:03 Quote
I have been using the 6-core AMD Istanbul variants for several months now and they have been quite nippy (2.4Ghz). Those 12 cores look snazzy but alas, the motherboard would have to be swapped to support DDR3.
Splynncryth 1st April 2010, 06:02 Quote
Quote:
Originally Posted by coolmiester
Made good reading and great to see Bit-Tech covering something for the dual CPU enthusiasts, especially as i just got a couple of X5650's :)

I'm currently running mine in an Intel S5520SC Shady Cove board running BIOS update 48 and getting in the region of 36014 in Cinebench R10 to your 32724 which is quite a large increase.

I'll run a few more of the benchmarks you ran and see how they compare.

I wonder if the difference is in the BIOS, I have some idea where the Intel BIOS stands in terms of errata, but I have no idea about Supermicro.
Are you using the integrated graphics on your shady cove board or a GPU? Might that be affecting the score?

Are the benchmarks free? I have have a few of the other Intel boards at work and it might be interesting to run the benchmarks on those.
Lizard 1st April 2010, 08:54 Quote
Quote:
Originally Posted by Splynncryth
Are the benchmarks free? I have have a few of the other Intel boards at work and it might be interesting to run the benchmarks on those.

Most of them are yes, you can find the download links on the benchmark description pages within the article.
phuzz 1st April 2010, 10:33 Quote
The AES support on the Intel chip doesn't mean that it's automatically encrypting data, it refers to a hardware accelerated AES implementation on the chip (it's usually done in software).
Anfield 1st April 2010, 13:08 Quote
How about a review of a 64 socket Nehalem EX system? http://www.intel.com/performance/server/xeon_mp/summary.htm?iid=products_xeon7000+body_benchmarks

I know I know, it won't happen as a) it sucks enough power to cause a blackout in the entire uk and b) its not exactly the kind of system aimed at the average bit-tech reader.:D, but I need my daily dose of geek pr0n.
Lizard 1st April 2010, 14:13 Quote
Quote:
Originally Posted by Anfield
How about a review of a 64 socket Nehalem EX system? [url]http://www.intel.com/performance/server/xeon_mp/summary.htm?iid=products_xeon7000+body_benchmarks

We can't do a 64-socket system, but we do have a quad-socket Nehalem EX review with 64-threads in the works :)
Turbotab 1st April 2010, 16:19 Quote
No database tests?, sadly many of these poor chips, will spend their lives as slaves to SQL / Oracle.
Anfield 1st April 2010, 23:12 Quote
Quote:
Originally Posted by Lizard
We can't do a 64-socket system, but we do have a quad-socket Nehalem EX review with 64-threads in the works :)

No worries, wasn't really expecting one;)
Lizard 2nd April 2010, 00:16 Quote
Quote:
Originally Posted by Turbotab
No database tests?, sadly many of these poor chips, will spend their lives as slaves to SQL / Oracle.

We're trying to develop a test at the moment, but it wasn't ready for these CPUs. Thanks for the reminder though, we do listen to our readers and are always trying to improve our reviews.
Splynncryth 2nd April 2010, 17:14 Quote
Quote:
Originally Posted by Lizard
Most of them are yes, you can find the download links on the benchmark description pages within the article.

I'll have to see about coming into work over the weekend and doing some unofficial testing.
leexgx 3rd April 2010, 05:22 Quote
can you post the frame times as well as the expected PPD for Folding@home
coolmiester 3rd April 2010, 09:49 Quote
Quote:
Originally Posted by Splynncryth
I wonder if the difference is in the BIOS, I have some idea where the Intel BIOS stands in terms of errata, but I have no idea about Supermicro.
Are you using the integrated graphics on your shady cove board or a GPU? Might that be affecting the score?

Are the benchmarks free? I have have a few of the other Intel boards at work and it might be interesting to run the benchmarks on those.

Sorry, missed that one.

I'm running a dual PCB GTX295 but i don't think that would effect a Cinebench R10 scores.

I have noticed quite a big fluctuation in score (as much as 2 to 3 K) though depending when the benchmark is run so that would account for the difference for sure. First run just after a restart yields much greater scores ;)
Lizard 3rd April 2010, 11:09 Quote
Quote:
Originally Posted by leexgx
can you post the frame times as well as the expected PPD for Folding@home

Sure thing, you'll have to wait till Tuesday when I'm back in the office though, as it's the long Easter weekend now.
Lizard 6th April 2010, 10:32 Quote
As promised here are the frame times for the bigadv folding client on the CPUs in this article:

2 x AMD Opteron 6174 = 18:02
2 x Intel Xeon X5650 = 18:10
2 x Intel Xeon W5580 = 23:00
2 x Intel Xeon X5482 = 32:00
Initialised 15th April 2010, 19:36 Quote
Can you post the scaling factor with the Cinebench data?

From what I've seen Hyper-Threading helps keep it linear as core count increases but this drops off after 4 cores with AMD's X6 getting 4.75 to 4.93 while i7 980X gets ~5.7.

Now that we are entrenched in the Core Wars it makes sense to look at the impact of Amdahl's law in massively multicore computational environments.
adlihajarat 9th July 2010, 21:40 Quote
I wonder if the code in those programs is optimized for the AMD microcode as it has to emulate the code when it runs the Intel code, which slows it down. I say this out of previous experience when I coded a program to test bench a coding scheme on different platforms running on AMD and Intel processors. The first run was using a code done for Intel, and the AMD was slower in that case, but when the same code was optimized for the AMD processor, things were there other way round, as the AMD won it clear.
pearl.of.wisdom 12th July 2010, 17:30 Quote
Surely, this is always the case, application code optimization has always had a massive impact on performance, that's why Intel has has been obsessed with preventing a fair playing field between the two; Intel will do anything, possibly short of arson and assassination, to keep developers from supporting AMD properly.
Tulatin 12th July 2010, 17:31 Quote
Quote:
Originally Posted by pearl.of.wisdom
Surely, this is always the case, application code optimization has always had a massive impact on performance, that's why Intel has has been obsessed with preventing a fair playing field between the two; Intel will do anything, possibly short of arson and assassination, to keep developers from supporting AMD properly.

I think you should change your username to Pearl of Bias.
pearl.of.wisdom 14th July 2010, 17:55 Quote
Quote:
Originally Posted by Tulatin
I think you should change your username to Pearl of Bias.

O Touche!, Dear Frog Sir [Or Madam], but please do be reasonable, coding optimizations do have such a terrible impact on performance; have you not heard the stories over the years of the fights that have been had? Wheather it may be Intel here, or Nvidia their, or AMD over that way.
And you left the dots out of my non deplume! You accursed amphibian!
Still, you're right a little bit; "To the last breath of my dying body, I will spit out my hate, my hate, too those demons at Intel Corp. Die! I tell you! Die!"
Tulatin 14th July 2010, 17:56 Quote
Quote:
Originally Posted by pearl.of.wisdom
O Touche!, Dear Frog Sir [Or Madam], but please do be reasonable, coding optimizations do have such a terrible impact on performance; have you not heard the stories over the years of the fights that have been had? Wheather it may be Intel here, or Nvidia their, or AMD over that way.
And you left the dots out of my non deplume! You accursed amphibian!
Still, you're right a little bit; "To the last breath of my dying body, I will spit out my hate, my hate, too those demons at Intel Corp. Die! I tell you! Die!"

While Intel is far and likely to be doing things like this, it's just a function of business. I'm fairly sure AMD does it - you don't have to demonize them to the point of running a lawnmower through a field of babies.
pearl.of.wisdom 16th July 2010, 17:47 Quote
Quote:
Originally Posted by Tulatin
While Intel is far and likely to be doing things like this, it's just a function of business. I'm fairly sure AMD does it - you don't have to demonize them to the point of running a lawnmower through a field of babies.
Well dear, Tulatin, I don't think I would, you know, eh lawnmower the baby thing. And true, all companies cheat to some degree if that think they can get away with it, God knows AMD is not some shinning knight, it's just, sigh, well, [a] Intel does is worse and [b] there seems to be so much either anti-AMD, or pro-Intel [whichever] feeling around. Anyway, I going off to hit my punchbag, it has a pic of Mr. Otellini on it.
Muad'Dib 28th July 2010, 17:23 Quote
Well, the conclusion is not only extremely biased in favour of Intel, but also fundamentally flawed.

Firstly, it is just absurd that 24 virtual cores at similar clock frequency and similar IPC could ever be faster than 24 real cores...
That said, most of the so-called "test suites" are mostly a case against themselves, rather than the processors. Clearly, if a test suite cannot fully utilize a computer (with 2 Opterons 6174), what is the point of having the computer go through it?

Anyway, I bet that most of the "overwhelming" results of Intel processors are actually the result of the ignominious Intel compiler...

Clearly, many commercial apps and their developers have slept on their laurels are simply not ready for the offering and challenge that the Opteron G34 socket system present for them.

Actually, it would be absurd to scathe AMD for gluing 2 processor dice, for Intel has done that for long; in fact, Intel wasn't able to produce a native quad-core CPU before Nehalem. All that matters is power consumption, which is, quite surprisingly, lower than Intel's Xeon, just check the numbers – remember that AMD's ACP means roughly the same as Intel's TDP.

Anyway, I disagree that the best explanation is IPC and clock frequency. I'd say it is the optimization – or, better said, the lack thereof – that makes and can explain the huge difference. The real difference only shows up with applications ready for the processors.

Besides, what WPrime does is nowhere near those "simplistic maths calculations", as alleged in the article.

And what the hell did you study, if such an obvious thing as the tangent iterative numerical method (also known as the Newton-Rhapson method) says nothing to you??!
Culinia 4th August 2010, 20:46 Quote
Where would one buy AMD Opteron 6174 and all the neccesary kit? I looked at the usuals i.e. Scan, OC, Ebuyer etc and none have them listed and don't sell the correct g34 motherboard.
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums