2x AMD Opteron 6282 SE benchmarks at SiSoftware

[BlueBlazer]

From SiSoftware >> Details for Device 2x AMD Opteron 6282 SE........

* Processor Arithmetic Benchmark: 231.971 GOPS
* Processor Multi-Media Benchmark: 585.983 Mpix/s
* Multi-Core Efficiency Benchmark: 21.773 GB/s
* Cryptography Benchmark: 4661 GB/s

Quite recent (September 7th, 2011). It seems that these are not Engineering Samples. No information on the stepping. 😉

 

[Riek]

From SiSoftware >> Details for Device 2x AMD Opteron 6282 SE........

* Processor Arithmetic Benchmark: 231.971 GOPS
* Processor Multi-Media Benchmark: 585.983 Mpix/s
* Multi-Core Efficiency Benchmark: 21.773 GB/s
* Cryptography Benchmark: 4661 GB/s

Quite recent (September 7th, 2011). It seems that these are not Engineering Samples. No information on the stepping. 😉

on the other hand there is also:

http://www.sisoftware.eu/rank2011d/...8feed3e5d0e7d6f082bf8fa9cca994a482f1ccfc&l=en from 6July where an ES scores 6967GB/s in the cryto test... which accidently is a higher clocked (+300MHz) 12core.. So their seems to be something seriously affecting scores in terms of BD... They seem to switch places in the other benchmarks.

 

[Abwx]

And what do you conclude , granted theses numbers are reals?..

 

[Riek]

And what do you conclude , granted theses numbers are reals?..

Well they are always the best ranking AMD system (core and frequency based).

Lets take the multimedia for example:
The closed AMD MC system is a little faster:

602.554 Mpix/s 4 / 48 2300 MHz

Which actually has 4 12cores @ 2,3GHz compared to two interlagos 16cores at 2GHz.

The next inline is qnother MC system which is a little slower
552.624 Mpix/s 4 / 48 2100 MHz


(compared to intel it still sucks in sisoft... but one can see this as a sisoft thing also.)



When you look at arithmitic the difference are alot smaller compared to the previous version.


4x AMD Opteron(tm) Processor 6176 SE (12C 2.3GHz, 1.8GHz IMC, 12x 512ko L2, 10Mo L3) 361.539 GOPS 4 / 48 2300 MHz Normal

4x AMD Opteron(tm) Processor 6136 (8C 2.4GHz, 1.8GHz IMC, 8x 512kB L2, 10MB L3) 255.566 GOPS 4 / 32 2394 MHz .

So it is still faster per core then MC, but not by that much.
That said the first quad system ranks 10th... so previous AMDs are doing pretty well in the arithmetic test

 

[Abwx]

Well they are always the best ranking AMD system. But there aren't alot of MC in those tests.

Right , but there is 48C MC systems , wich should have , according
to AMD , more or less the same performance as this 32C BD.

 

[inf64]

Nice find. At what clock speed is that thing running? If AMD follows MC naming scheme it should be 2.5Ghz. Sisoft detects it as 2.0Ghz though. Anyway,for comparison purpose:
2x 12C Opteron @ 2.5Ghz
2x 12C Opteron @ 2.3Ghz

If this sample is running at 2.5Ghz then FP performance is around 35% better per core and per clock than MC. Kinda conflicting with what we see from Zambezi and Cinebench/wprime,ain't it?
Integer throughput is a bit lower per clock though. This is if we presume it runs at the same clock as MC. If it runs at 2Ghz then it is 7.4% faster per core in this trhoughput workload(and if 80% sclaing thing is true for integer,then each core is 1.074x1.2~=1.3x or 30% faster than MC core). Cryptography is much lower than what that ES 16C Interlagos got in July,around 50% lower. Something is affecting the scores.It could be that every sample has been "marked" in some way,so AMD knows exactly who leaked what 🙂.

 

[Riek]

Nice find. At what clock speed is that thing running? If AMD follows MC naming scheme it should be 2.5Ghz. Sisoft detects it as 2.0Ghz though. Anyway,for comparison purpose:
2x 12C Opteron @ 2.5Ghz
2x 12C Opteron @ 2.3Ghz

If this sample is running at 2.5Ghz then FP performance is around 35% better per core and per clock than MC. Kinda conflicting with what we see from Zambezi and Cinebench/wprime,ain't it?
Integer throughput is a bit lower per clock though. This is if we presume it runs at the same clock as MC. If it runs at 2Ghz then it is 7.4% faster per core in this trhoughput workload(and if 80% sclaing thing is true for integer,then each core is 1.074x1.2~=1.3x or 30% faster than MC core). Cryptography is much lower than what that ES 16C Interlagos got in July,around 50% lower. Something is affecting the scores.It could be that every sample has been "marked" in some way,so AMD knows exactly who leaked what 🙂.

It is really hard to determine performance advantage /clock because we have no id how Sisoft scales with cores. + we don't know the clockspeed of interlagos. We know there is a turboboost of 500MHz that can be active on all cores depending on TDP. Given 500MHz is 25% more, this is pretty significant detail 🙂.

 

[inf64]

As one user at XS noticed,maybe this 6282SE was ran at TDP limited clock (via new feature called power cap). This way it stays always below certain predetermined TDP limit (say 75W). This would explain odd clock that is listed for a supposedly top Interlagos model (since 6276 runs at 2.3Ghz base).

 

[wlee15]

The 6282SE looks pretty competitive to the Sandy Bridge-EP samples in the arithmetic and multi-media benchmarks.

 

[JFAMD]

Doesn't look real to me. Seriously guys.

 

[LOL_Wut_Axel]

SiSoft Sandra sucks. It's synthetic garbage that gives preference to Intel's CPUs.

Also, John just said he doesn't think it's real.

 

[BlueBlazer]

Doesn't look real to me. Seriously guys.

Are you're implying SiSoftware published "fake" results? :hmm: Perhaps you could show us some "real" results as proof, and if possible also some SPECint and SPECfp ones (since these are server CPUs). 😀

 

[jhu]

Synthetic benchmarks can be misleading. What really matters is performance/price for real world applications and what you're willing to spend. For example, I use Povray for rendering and occasionally play Guildwars. Now, according to this the AMD Phenom x6 1100T sits between the Core i7 2600k and Core i5 2500K in terms of rendering. However, the AMD chip is $110 less than the i7 and $20 less than the i5. For my money, I'd rather get the x6 1100: it's quite a bit slower than i7 but also much cheaper, yet is both cheaper and faster than the i5.

If I were to only base my decisions on FPU benchmarks (ie SPEC), I would have gotten an Itanium2 system (they're reasonably priced now on ebay). However, on closer inspection, Itanium doesn't actually perform very well with Povray.

 

[amenx]

Synthetic benchmarks can be misleading. What really matters is performance/price for real world applications and what you're willing to spend. For example, I use Povray for rendering and occasionally play Guildwars. Now, according to this the AMD Phenom x6 1100T sits between the Core i7 2600k and Core i5 2500K in terms of rendering. However, the AMD chip is $110 less than the i7 and $20 less than the i5. For my money, I'd rather get the x6 1100: it's quite a bit slower than i7 but also much cheaper, yet is both cheaper and faster than the i5..

One can always cherry pick a benchmark or two where a product, CPU or GPU, does unusually well vs the competition. But when you look at the whole picture, we see something entirely different.

http://www.anandtech.com/bench/Product/203?vs=288

There are even several benches where an i5-750 leaves the 1100t behind. D:

 

[BlueBlazer]

Synthetic benchmarks can be misleading.

They usually and primarily measure throughput (how much instructions executed, memory read/write and FP calculations that can be crammed into the CPU). Synthethics are still used in reviews nowadays for comparing CPUs. It can be used to point to issues with the CPU (or software) when comparing with other real world benchmarks. :hmm:

What really matters is performance/price for real world applications and what you're willing to spend. For example, I use Povray for rendering and occasionally play Guildwars. Now, according to this the AMD Phenom x6 1100T sits between the Core i7 2600k and Core i5 2500K in terms of rendering. However, the AMD chip is $110 less than the i7 and $20 less than the i5. For my money, I'd rather get the x6 1100: it's quite a bit slower than i7 but also much cheaper, yet is both cheaper and faster than the i5.

Those desktop applications (and desktop CPUs). The Core i7 2600K is roughly 25% faster, roughly this could cut down long rendering times (more if rendering multiple frames). For people serious on 3D rendering (for work) would be seriously considering these Core i7 CPUs (and dual socket machines). Unfortunately in your case, affordability is the reason you went for Phenom II X6 1100T. Anyway, server CPUs costs much more than desktop ones. 😛

If I were to only base my decisions on FPU benchmarks (ie SPEC), I would have gotten an Itanium2 system (they're reasonably priced now on ebay). However, on closer inspection, Itanium doesn't actually perform very well with Povray.

Itanium is a speciality product (for certain markets like Oracle database). Its hardware architecture or available compiler is may not be suited for Povray. SPEC is relevant in servers (and HPC). Often purchasing decisions can be based on SPEC benchmarks. Just like when Opterons appeared, they are suited for high FP thoughput (compared to FSB based Xeons) partly due to advantages of integrated memory controller, and was widely adopted in HPC (and supercomputers, although this is later challenged by Intel's Nehalem and its next generations). Also x86 architecture based compilers has much more development and maturity. Speaking of which, I still haven't seen any official SPEC benchmarks for Interlagos. :hmm:

 

[jhu]

One can always cherry pick a benchmark or two where a product, CPU or GPU, does unusually well vs the competition. But when you look at the whole picture, we see something entirely different.

http://www.anandtech.com/bench/Product/203?vs=288

There are even several benches where an i5-750 leaves the 1100t behind. D:

Which is why I'm glad anandtech uses Povray as one of their benchmarks because I use that program. If I had based my purchase on 3dsmax benchmarks under the impression that it would be representative of Povray (not all websites test with Povray unfortunately), I would have ended up with a slower and more expensive processor. Now if only they'd test with RC3 instead of beta23, which is obsolete and can't be downloaded anymore.

 

[jhu]

Itanium is a speciality product (for certain markets like Oracle database). Its hardware architecture or available compiler is may not be suited for Povray. SPEC is relevant in servers (and HPC). Often purchasing decisions can be based on SPEC benchmarks. Just like when Opterons appeared, they are suited for high FP thoughput (compared to FSB based Xeons) partly due to advantages of integrated memory controller, and was widely adopted in HPC (and supercomputers, although this is later challenged by Intel's Nehalem and its next generations). Also x86 architecture based compilers has much more development and maturity. Speaking of which, I still haven't seen any official SPEC benchmarks for Interlagos. :hmm:

There was an article a few months back (in Arstechnica I think) that covered the Itanium. I think performance/price doesn't justify it (and hasn't for it's entire existance). Oracle has already dropped support for Itanium. Look at top500.org; there are only 5 Itanium systems on the list and none in the top 10. For any given amount of money, you can purchase much more computing power with x86 than with Itanium.

 

[radaja]

heres x2 opteron 6220 8c at 2.75GHz from sisoft,although cpu world has them listed as 3.0Ghz cpu's.its from sept 7th 2011

2x8c opteron 6220@2.75Ghz

Scores
Processor Arithmetic Benchmark
score-138.538 GOPS

Processor Multi-Media Benchmark
score-315.004 Mpix/s

 

[BlueBlazer]

heres x2 opteron 6220 8c at 2.75GHz from sisoft,although cpu world has them listed as 3.0Ghz cpu's.its from sept 7th 2011

2x8c opteron 6220@2.75Ghz

Seems to be from Chinese machine, possibly from a motherboard manufacturer testing and validation (speculation). :hmm:

Scores
Processor Arithmetic Benchmark
score-138.538 GOPS

Processor Multi-Media Benchmark
score-315.004 Mpix/s

The scores seems to get lower as the clock frequency increases (when compared, taking into account the number of cores and clock speed). :\

 

[inf64]

Ok I think I kinda figured out what is going on. Here. Maybe it's a bit complicated but fits with corrected HPC increases AMD now states for Interlagos(35% higher throughput in HPC versus top bin MC). Tell me what you think.

 

[Idontcare]

Ok I think I kinda figured out what is going on. Here. Maybe it's a bit complicated but fits with corrected HPC increases AMD now states for Interlagos(35% higher throughput in HPC versus top bin MC). Tell me what you think.

Seems reasonable to me. I'm gonna be pissed if post 47 turns out to be the case though.

 

[BlueBlazer]

Seems reasonable to me. I'm gonna be pissed if post 47 turns out to be the case though.

I think that will depend on application tested. Remember David Kanter's analysis? It might vary, whether its server or desktop type applications. In servers, specific compiler optimizations can used for tuning. However in desktop, applications are already pre-compiled (fixed). Thus would be interesting to see how this turns out. :hmm:

 

[Idontcare]

I think that will depend on application tested. Remember David Kanter's analysis? It might vary, whether its server or desktop type applications. In servers, specific compiler optimizations can used for tuning. However in desktop, applications are already pre-compiled (fixed). Thus would be interesting to see how this turns out. :hmm:

Well if that distinguishing factor turns out to be true, then I think we all are justified to ask ourselves and AMD what good was to ever come of having a server guy go to desktop-facing forums to espouse the position that "bulldozer IPC does not decrease" if in fact there was a huge and relevant caveat to that point which was "unless you happen to be a desktop-oriented customer here on this desktop-oriented forum, in which case your desktop-based apps will likely turn out to experience lower IPC".

It does no one any good to speak in terms of inclusives while hiding behind caveats that form a stronghold of exclusives.

 

[Arachnotronic]

I'm still kinda holding out hope that Bulldozer is good...

 

[podspi]

Well if that distinguishing factor turns out to be true, then I think we all are justified to ask ourselves and AMD what good was to ever come of having a server guy go to desktop-facing forums to espouse the position that "bulldozer IPC does not decrease" if in fact there was a huge and relevant caveat to that point which was "unless you happen to be a desktop-oriented customer here on this desktop-oriented forum, in which case your desktop-based apps will likely turn out to experience lower IPC".

It does no one any good to speak in terms of inclusives while hiding behind caveats that form a stronghold of exclusives.

I've always been under the impression that JF hangs around here on his own time, kind of like me 😎

(Sometimes for a very liberal interpretation of "my time)

If average IPC turns out to actually decrease, which I doubt, then I would probably argue that JF was mistaken/given bad info, since last time I checked he was not actually an engineer on the project.

But seriously, lower IPC? Given how far behind they already are in IPC, that would be crazy, unless they're planning another new arch for client-products. AMD can't expect Trinity to hit ~ 4ghz in a laptop... I think...