bit-tech.net

AMD Opteron 6174 vs Intel Xeon X5650 Review

Comments 1 to 25 of 50

Reply
Tulatin 31st March 2010, 08:46 Quote
I'm hoping this line isn't a misprint
6128 2GHz 8 80W 1,333, 1066, 800 $266
Though the motherboards are FAR from economical (in regards to a specialty case requirement, and high set up costs), 8 physical cores at 2GHz a piece is a rather INSANE value for $266.
Lizard 31st March 2010, 08:55 Quote
No, it's not a misprint, but don't forget that $266 is the list price, so what you end up paying (depending on where you buy it) won't be exactly the same as this. It's a great price even so :)
confusis 31st March 2010, 09:23 Quote
looks like folding has a new king... might be time to save for a amd 6174setup!
barndoor101 31st March 2010, 09:27 Quote
you got the Xeon table wrong - the logical cores column
500mph 31st March 2010, 09:49 Quote
Was I the only one who thought the Picture was stretched when they first saw the Opteron?
Tulatin 31st March 2010, 09:54 Quote
Quote:
Originally Posted by Lizard
No, it's not a misprint, but don't forget that $266 is the list price, so what you end up paying (depending on where you buy it) won't be exactly the same as this. It's a great price even so :)

Even if it hits the deck for $299 retail, it's still a fantastic deal if you think about it. It would really open the door to rather ridiculous performance, when you consider that you could probably build a 16 core, 32GB of ram rig for a shade under 5k.

I'm curious as to those motherboards, though. Judging by the haphazardly placed pci-e slots and giant motherboard gaps, I assume there's I/O risers, or?
coolmiester 31st March 2010, 09:54 Quote
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.
Lizard 31st March 2010, 10:09 Quote
Quote:
Originally Posted by barndoor101
you got the Xeon table wrong - the logical cores column

What do you think is incorrect in the table?
Tulatin 31st March 2010, 10:14 Quote
Quote:
Originally Posted by Lizard
What do you think is incorrect in the table?

It may be difficult to utilize a CPU with 0 logical cores.

L5609 1.86GHz 4 0 40W 1,066, 800 $440
Kúsař 31st March 2010, 10:15 Quote
With this many threads available I'm curious how would two or more "heavy" applications run on these...
Lizard 31st March 2010, 10:16 Quote
Quote:
Originally Posted by Tulatin
Quote:
Originally Posted by Lizard
What do you think is incorrect in the table?

It may be difficult to utilize a CPU with 0 logical cores.

L5609 1.86GHz 4 0 40W 1,066, 800 $440

No, the table is correct. This CPU does have 4 physical cores but doesn't support Hyper-Threading so doesn't have any logical cores.
Tulatin 31st March 2010, 10:22 Quote
Quote:
Originally Posted by Lizard
No, the table is correct. This CPU does have 4 physical cores but doesn't support Hyper-Threading so doesn't have any logical cores.

Makes sense. Can you add an * to it in the article, and clarify this to deal with further confusion?
barndoor101 31st March 2010, 10:27 Quote
Quote:
Originally Posted by Lizard
No, the table is correct. This CPU does have 4 physical cores but doesn't support Hyper-Threading so doesn't have any logical cores.

ahhhhhhh right, you were doing it in addition as opposed to giving the total - normal nomenclature is to give the total physical cores then the total logical cores, that way you can easily tell if the CPU has HT.

a single physical core without HT is still a single logical core ;)
Lizard 31st March 2010, 10:29 Quote
Quote:
Originally Posted by Tulatin
Quote:
Originally Posted by Lizard
No, the table is correct. This CPU does have 4 physical cores but doesn't support Hyper-Threading so doesn't have any logical cores.

Makes sense. Can you add an * to it in the article, and clarify this to deal with further confusion?

Good idea, I've added an addendum underneath the table to hopefully avoid any confusion over this issue.
Doomah 31st March 2010, 11:13 Quote
About the X5677. The reason it's higher clocked but less cores is simple: Some server software is 'pay per core'. So if your server has less cores, you pay less for the software. So instead of taking a slower 6 core, you actually save money by taking a faster quadcore.
barndoor101 31st March 2010, 11:16 Quote
Quote:
Originally Posted by Doomah
About the X5677. The reason it's higher clocked but less cores is simple: Some server software is 'pay per core'. So if your server has less cores, you pay less for the software. So instead of taking a slower 6 core, you actually save money by taking a faster quadcore.

its probably more to do with certain applications not scaling above 8 threads - so having 8 faster threads is more desirable than ahving 6 slower ones.
unclean 31st March 2010, 12:45 Quote
Guys, the explanation of why the Opteron is so much faster in Cinebench R11 is flawed. Although there is some element of truth regarding the lower clocked cores, as there is more overhead associated with actually executing each thread, that can be said of R11 too. The difference is really due to the 16 thread limit present in Cinebench R10.

I don't know, it seems like nowadays there's an element of thorough analysis that's missing and results are rushed online.

Same for the 470 & 480 reviews. Arguably, the 480 *is* the fastest single card, not by much of a margin, granted. However, to give the 470 lower ratings (mainly regarding value) than the 480 seems absurd, as it clearly represents much better value for money.

I feel like a parent saying "I'm not angry, just disappointed"...
venommachine 31st March 2010, 13:14 Quote
Post by; Tulatin:

I'm curious as to those motherboards, though. Judging by the haphazardly placed pci-e slots and giant motherboard gaps, I assume there's I/O risers, or?

The Supermicro UIO boards come with various different PCI-E Risers, to go in either 1U or 2U form factors. Supermicro of course have DP Workstation solutions available also for mid-tower or workstation form factors for personal supercomputing.

Boston Limited, where the test kit came from, are the guys who know all :)

JW
Lizard 31st March 2010, 13:18 Quote
Quote:
Originally Posted by unclean
Guys, the explanation of why the Opteron is so much faster in Cinebench R11 is flawed. Although there is some element of truth regarding the lower clocked cores, as there is more overhead associated with actually executing each thread, that can be said of R11 too. The difference is really due to the 16 thread limit present in Cinebench R10.

Good point, sorry about that - I've been trying to write the Opteron 6000-series, Xeon 5600-series and forthcoming Xeon 7500-series reviews together, and had overlooked the thread limit in Cinebench R10.
Quote:
Originally Posted by unclean
Same for the 470 & 480 reviews. Arguably, the 480 *is* the fastest single card, not by much of a margin, granted. However, to give the 470 lower ratings (mainly regarding value) than the 480 seems absurd, as it clearly represents much better value for money.

This isn't the right place to discuss this, as there's two dedicated threads for the GTX 470 and 480, but just because something is cheaper doesn't mean it's better value for money.

Value for money is an indication of the balance between performance/features/price - not just an indication of how cheap a product is.
javaman 31st March 2010, 13:38 Quote
So in including ram and mobo, how much would a quad socket settup cost? Would that displace GPU's for folding?
venommachine 31st March 2010, 13:45 Quote
Quad socket (or MP) pricing is available, contact Boston Limited - www.boston.co.uk or sales@boston.co.uk for more information.

Ping me your number and I can call you to discuss if you would like.

I wouldn't say it would displace GPU's for folding, but that is a killer score from the Opteron's! - Fermi GPU's will be available soon in some pretty meaty flavours - one to watch out for!!

Potentially if you have an interesting project, we could invite you in to use the Boston Labs for your benchmarking :)
Lizard 31st March 2010, 15:33 Quote
Quote:
Originally Posted by javaman
So in including ram and mobo, how much would a quad socket settup cost? Would that displace GPU's for folding?

If you're really interested in quad-socket then I'd recommend waiting for our review of Intel's 4P Nehalem EX product in the next few days.
F40 31st March 2010, 18:47 Quote
Magny-Cours that's a race circuit in France, is it not!!!
comdot 31st March 2010, 19:25 Quote
Before I start I do understand bit-tech's focus is the consumer market..

I find it very odd that virtualisation isn't even mentioned in the article? As you say in the piece, very few single applications can take advantage of cores. The ideal application for these chips and the application the majority of them will blt probaend up in is virtualisation, I would love to see some vm-ware or hyper-v benchmarks included when testing products like these!!

I'm sure I'm not unusual in being someone who works in IT yet also enjoys the journalism and content at bit-tech / custom-pc.

comdot
Lizard 31st March 2010, 21:33 Quote
Thanks for your feedback comdot.

To be honest we've never been asked to do any virtualisation tests before so that's why there aren't any included in this article. However, I'm sure it's something we can look at including in future articles like this.
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