This can't be good for Nvidia if it's true. :(
linkQuote:
REMEMBER THE TRIUMPHANT WIN for Fermi at the Oak Ridge National Laboratory that Nvidia heavily touted at its GTC conference keynote? The supercomputer project was just killed for power reasons. Fermi power reasons. Whoops.
tsubame is still in line for fermi and they have a 1 megawatt power budget.:up:
that sucks it was cancelled. it could have done a lot of research.
So fermis use a lot of power? That's not usually an issue with supercomputers unless the perf/W is also low.
190TDP for the 3gb base model quadro
My guess is they did not want to wait until early Q2 2010....
im sure they will do even more research with better cardsQuote:
Originally Posted by Chumbucket843
It's all good. I know of a nice place in 'Longhorn Country' that is under NDA developing a visual setup with Fermi as we speak :)
You mean the ones that don't exist?Quote:
Originally Posted by Machinus
ATIs cards seem pretty real to me.Quote:
Originally Posted by 003
You said "better cards" though, which rules out any current ATI cards. Future card I'm not sure.
Yes, because assuming that a card will be better than one that is not released yet it a logical thought process.
Well CUDA on GT200 already makes the OpenCL implementation on any ATI card look broken by comparison and CUDA is going to be further improved on Fermi so it would logical to assume it is better yet.
dude ... read this first
http://www.xtremesystems.org/forums/...d.php?t=241120
I want them to use ati cards instead i'd find it hilarious..
anyway this foreshadowing high power req's I reckon.. they have to do something with power anyways they can ONLY have 300w right :S cos of pcie spec? but yeah :S external power would be the way forward.... I wish we had somethign to more efficiently cool these cards. (by these cards i mean ALL modern cards)
Interesting, some sources have stated it wasn't cancelled but rather expanded and given a new name. That could be why Charlie heard it was cancelled.
I highly suggest you wait for official confirmation from ORNL on this one. They are usually very up front with their press announcements even if they are cancelling a project.
What sources are these? I'd love to see them and I hate Charlie! :mad:Quote:
Originally Posted by SKYMTL
No, ATIs current cards will beat anything nVidia releases in the next year for power efficiency. And it sounds like 5890 will actually be faster than fermi anyway, and come out sooner. So no, nVidia loses no matter which way you look at it.Quote:
Originally Posted by 003
You must know something I don't considering there is zero official information on power and performance on the 5890 and the power and performance of the GeForce based Fermi. Even I don't have that specific information with the NDA info.Quote:
Originally Posted by Machinus
Are you guys seriously talking about power efficiency? The competition is CPUs and the most power inefficient GPU is still far more efficient than the best CPU.
In either case Fermi has no competition. HD 58xx's gaming performance means squat in an environment where Fermi's ECC support and cache hierarchy are a big deal.
actually you would be wrong. even if they were they would not even be a scalable or reliable option. fermi will tear apart r800 in double precision too.Quote:
Originally Posted by Machinus
Apparently I do know something you don't. It is common knowledge that the 5800 series has the best power efficiency in several generations and 5890 is not going to be any different. At the moment ATI has shipped about a million of these cards, while nVidia is still selling the same 8800 GTX from ten years ago with different plastic on it (I'm not counting the fake fermi boards that were at the development conference, also with different plastic on them).Quote:
Originally Posted by 003
If nVidia can ever figure out how to design chips at 40nm so that they actually work when they come out of the factory, they can sell a product...until then ATI is going to be planning exactly how they are going to spoil the launch day. If you read any sort of news about the GPU industry you would have sold your stock in nVidia desktop graphics by now...I'd suggest getting out before it's too late
LMAO, and what do you base that statement on? 003 and his fans are hilarious!!! :ROTF:Quote:
Originally Posted by Chumbucket843
Im not saying Fermi can not beat ATI's cards, but hell... nobody knows what Fermi will be like ATM.
nvidia already announced speed of tesla. so thank you for the insult.:shakes: if you dont believe me go on their website.Quote:
Originally Posted by nullface
the scalability and reliablity is from ecc which no other gpu has.
So what exactly is the TDP of the high end Fermi based GeForce? And what about 3D performance? Because without that information you can not logically make the claims you are trying to assert.Quote:
Originally Posted by Machinus
OpenCL implementation on ATI is garbage. Talk to any programmer who has done work with ATI and OpenCL and Nvidia CUDA. CUDA is much more straightforward and intuitive and supported, and better performing as well (more direct access to the GPU). And it's about to get a lot better with Fermi. Plus Fermi has ECC which will be the deciding factor for any organizations and institutions who are serious about GPGPU.Quote:
Originally Posted by nullface
And GeForce wont have it ether (Quadro only), or that is what the "The GT300/Fermi Thread" tells me, again quotes with no source.Quote:
Originally Posted by Chumbucket843
Show me an actual review/benchmark and i will believe everything, till then i will just laugh my ass to hell and back every time someone hypes Fermi or even the 5800 over Fermi.
It is a very good thing that GeForce won't have ECC. It is a significant performance hit. Especially because it is 2-bit ECC.
Sources are sources. Anonymous stays anonymous. Charlie doesn't burn his so why should anyone else in this industry burn theirs? That's the way investigative journalism works. ;)Quote:
Originally Posted by 003
As I said: I would expect an announcement in the next few weeks for better or worse.
As for which card is "better" for these scenarios, I would expect Fermi-based Tesla cards will trounce anything ATI can produce for HPC specific applications. Fermi was tailor made for HPCs and as such incorporates technologies that certain sectors have been looking for and the competition has no hope of delivering anytime soon.
ECC is not only a perf hit, Nvidia's implementation uses up a considerable amount of space too.
Isn't it actually the HD 5800 that has no competition? I don't think you can come to any conclusions until the card gets to the store shelves.Quote:
Originally Posted by trinibwoy
You are talking DX11. This thread is about HPC applications where ATI has little to no ongoing support of any kind.Quote:
Originally Posted by flippin_waffles
It sounds like you lack some basic logical reasoning abilities. There is plenty of information about how these firms design chips, and the resulting power specs...maybe you are in denial because nVidia has been selling :banana::banana::banana::banana: in a box for the last year? I suggest you go read the tech specs of RV800 (can you find those on your own?) and then read about the JumboDie TM design scheme that nVidia thinks is going to work (and is producing record low yields). A basic understanding of chip physics will tell you that 5890 is definitely going to beat whatever nVidia comes out with in terms of performance/watt. With such crappy yields and ATIs headstart (and opportunity to die shrink and cut prices) they are going to be also losing in performance/dollar. At this point the only thing that nVidia could possibly win at is total XOPS, which considering the differing technologies may be better for WCG etc. at performance/card, but that is a relatively useless metric in any market that matters...Quote:
Originally Posted by 003
I take a nap and Fermi threads start sprouting up like weeds on my lawn.
Take the discussion here please -> http://www.xtremesystems.org/forums/...d.php?t=241120
Regarding the original Fermi thread that was moved to the Nvidia section, I didn't know it existed else i would have moved it back.