AMD Releases Interlagos Opterons With Up to 16 Cores

[acadia11]

[citation][nom]shqtth[/nom]I would rather still use a magny cours cpu as it would be true cpu cores and not ones with screwed up fpu units[/citation]

They need to do something seriously aobut the woefully lacking FPU power of bulldozer.

 

[billybobser]

News just in.


Ah forget it.


I don't particularly care, intel's 2500k is very reasonably priced for what you get from it, it's just I'd rather it was cheaper.

From a server point of view, I don't even have a server to comment.

 

[digiex]

Opteron is very much alive in HPC environment.

 

[shqtth]

[citation][nom]acadia11[/nom]They need to do something seriously aobut the woefully lacking FPU power of bulldozer.[/citation]
Agreed.

If the fpu is shared, they should of somehow figured out how to increase the amount of fpu instructions that can be executed/fpu power/sse power. It is possible everytime the fpu needs to change hand from cpu to cpu within a module there is a stall, as things have to be finished first or work has to be saved. fpu wa snever designed to take orders from two cpus at once. two threads at once on an fpu is hard due to the stack structure of the fpu and the register structure of 3dnow/sse2. So i guess work would have to be saved somewhere for change in cpu, unless the complex instructions needs to be finished first (upt to so many) before cpus can be changed, adding a large latency.

 

[lunchbox4k]

[citation][nom]shqtth[/nom]Agreed.If the fpu is shared, they should of somehow figured out how to increase the amount of fpu instructions that can be executed/fpu power/sse power. It is possible everytime the fpu needs to change hand from cpu to cpu within a module there is a stall, as things have to be finished first or work has to be saved. fpu wa snever designed to take orders from two cpus at once. two threads at once on an fpu is hard due to the stack structure of the fpu and the register structure of 3dnow/sse2. So i guess work would have to be saved somewhere for change in cpu, unless the complex instructions needs to be finished first (upt to so many) before cpus can be changed, adding a large latency.[/citation] Ya AMD has stated that the current software is not opt amazed to use the new Bulldozer architecture. So threads that may need to share FP math would get put on cores that don't share a FP module and would need L3 cache to transfer making things a complete headache. Read this for more on how Windows 8 will fix this.

 

[STravis]

"Bulldozer-based Opterons are offer more efficient economics "

Are you getting paid by the word?

WRT to the AMD offering, I've always liked Opterons and this package seems like a winner. Yeah, maybe Bulldozer isn't a gamer's chip, but based on the push for cloud computing, maybe it was never intended to be one.

 

[lunchbox4k]

Actul Review Link

 

[saturnus]

This is not about Bulldozer vs. Sandy Bridge but about Opteron vs. Xeon which are server processors. In the server world, all that matters is performance per watt and performance per $. Everything else is more or less irrelevant.

 

[panders4]

I'm glad they chose to make the FPU shared. Any truely floating point heavy job should be offloaded to the GPU anyway. At least in an ideal world. That seems to be the way GPUs are going anyway.

 

[BulkZerker]

[citation][nom]lenell86[/nom]dear wintermint, stfu, bye, clueless moron[/citation]

lenell86 what you've just said is one of the most insanely idiotic things I have ever heard. At no point in your rambling, incoherent response were you even close to anything that could be considered a rational thought. Everyone in this room is now dumber for having listened to it. I award you no points, and may God have mercy on your soul.

 

[CaedenV]

[citation][nom]otacon72[/nom]Um the Xeon X5670 has 6 cores. Intel should make an X5670 with 16 cores and blow the Opteron 6276 out of the water just to shut AMD up.[/citation]
In fact the SB-E Xeon chips should be coming Q1, and will have 4, 6, and 8 core setups, with HT.
However I think that the bulldozer architecture will do well in this segment as more cores will be used more of the time, and they will be cheaper than the SB-E chips per performance. Best of luck to them, they could use a boost right now to get them through this rough patch, but I think they are on the right track.

 

[shqtth]

[citation][nom]lunchbox4k[/nom]Ya AMD has stated that the current software is not opt amazed to use the new Bulldozer architecture. So threads that may need to share FP math would get put on cores that don't share a FP module and would need L3 cache to transfer making things a complete headache. Read this for more on how Windows 8 will fix this.[/citation]

That is the biggest mistake. THy should of never designed for something that might work better for windows 8 or design something fo only windows 8 in mind. As it is now, just as much people with xp also run windows 7 (winows 7 users are starting to equal xp users). The fast is a lot of people run legacy OS. Only a few will bennifit from windows 8. I don't like this 'suffer now, and the future may get better' attitude.

Why couldn't they just trick the OS, to make it think each module is its own numa cpu with its own memory. that way, there wouldnt be as bad as a thread problem. that might help the cache problem too indirectly, as each cache wont have to cache as much ram.

 

[serendipiti]

Well, for those who read the Bulldozer review, the remarkable benchmarks for a server environment were the benchmarks where BD shone... The rest is marketing hype, ok. But the comparison 12 threads vs 16 BD cores will be fair or not depending on costs... until intel releases its news Xeons or moves prices... BD its not a bad architecture, its an evolutive technology and one step ahead, maybe people who waited for a i5 killer got disappointed. For the Desktop, I think Fussion chips are more relevant (SoC makes more sense for a desktop than a multichip configuration with 32 cores...).

 

[fazers_on_stun]

All very nice, but the Xeon X5670 is 20months old and Sandy Bridge Xeons are right around the corner.

True. The 6-core 3960X that was officially released today is a die-harvested 8-core Sandy Bridge EP part, and according to the reviews it blows the Bulldozer 8150 out of the water, 6 cores to 8 cores. So a true 8-core SB should be even more lopsided in comparison.

 

[MasterMace]

[citation][nom]otacon72[/nom]Um the Xeon X5670 has 6 cores. Intel should make an X5670 with 16 cores and blow the Opteron 6276 out of the water just to shut AMD up.[/citation]

And what? Charge $8,000 for it? Cuz they already have an 8 core nehalem at ~$4k

 

[bfstev]

judging by the comments, not many people seem to know much about the bulldozer processor other than "it sucks!". Bulldozer should do very well in the server space as it does seem to be more aimed at it. They need to come out with a more gaming/consumer orriented part or just leave the consumer market completely(i really hope they dont but thier inability to compete except on price is not very reassuring.

 

[joytech22]

[citation][nom]otacon72[/nom]Um the Xeon X5670 has 6 cores. Intel should make an X5670 with 16 cores and blow the Opteron 6276 out of the water just to shut AMD up.[/citation]

They will and probably have, but then you'd be paying out the butt for it.

 

[Guest]

I BELIEVE AMD

 

[enzo matrix]

[citation][nom]lenell86[/nom]dear wintermint, stfu, bye, clueless moron[/citation]
Let me put what you said together using proper grammar and letter format:

Dear Wintermint,
STFU
Bye,
Clueless Moron

So as we can see, the "stfu" was from someone who seems very aware that they are a clueless moron. So we can all stop insulting him because he already knows it.

 

[zak_mckraken]

^ lol

 

[Guest]

FP server load is all moving to GPUs anyway, and cpus get slaughtered compared to a gpu. So why bother trying to beef up the FPU in the server world, it makes no sense.

Going after integer performance and making a beefy server cpu was the right choice.

That said the bulldozer is disappointing as a desktop chip(i wont be buying a first gen, who knows about future gens until they are released). But, this article isnt about desktop, its about servers.

 

[ohim]

[citation][nom]rahulkadukar[/nom]AMD, nobody runs Linpack on their machines, do some real world tests and get humiliated[/citation]
like using itunes to transcode mp3, play sysmark and other benchmarking problems ... sureeeeeeeee and yes ... opterons to play some stupid game no ?

 

[gm0n3y]

Tentative props to AMD for FINALLY pulling a concrete success off (independent reviews notwithstanding).

 

[wiyosaya]

16 cores and hits less than 100% better than an X5670? Trust me, I'm an AMD fanboy, 😛, but this does not sound good. If the processor cost is significantly lower than the X5670, then AMD may have something. Too bad there is nothing about cost in the article.

Tom's how about a good analysis and benchmarks article on the chip?

 

[hetneo]

This is such a huge blunder on AMD's part. This is supposed to be their flagship server CPU but they stacked it against almost 2 years old Intel one, which is far from being Intel's flagship, not even the strongest in it's series. For those who didn't know or care to know, there's 6 more CPUs in X5xxx series, plus there Intel will soon launch E/X6xxx, L/E/X7xxx, E3 and E7 lines. As AMD fan I must say that Bulldozer reminds me of Intel's Netburst and that I hope they will be able to reinvent them self in shorter time than Intel needed.