Well, if the National SuperComputer Center in Tianjin/NUDT can be #5 in the Top500 using RV770s, it has to be doable...
NVIDIA Fermi: Architecture discussion
- Thread starter Rys
- Start date
Well, if the National SuperComputer Center in Tianjin/NUDT can be #5 in the Top500 using RV770s, it has to be doable...
Well whatever it is it would have to outdo the allegedly disappointing version of Fermi.
I will have to defend charlie on this one. There is no way windows is used in supercomputers or development for them, so visual studio is irrelevant.
Try eclipse or good old emacs/vi.
That's ignoring the gist of the post which was to highlight the deficiency on the software side of things. That's the case regardless of operating system or platform.
Until we have reliable indications what the frequencies of the highest end GF100 will be, quite a few performance aspects are still in the air. I for one hope that they can at least hit equivalent GTX285 frequencies, otherwise anything significantly lower than that might end up underwhelming for a new technology generation and even more so a 3.08b transistor chip.
Frequencies aside what is still a quite big and important question mark, are the 3D capabilities of GF100 and if there have been any worthwhile changes to raise efficiency. So far all we know are the HPC/Fermi capabilities which are hardly any serious indication for the former.
Frequencies aside what is still a quite big and important question mark, are the 3D capabilities of GF100 and if there have been any worthwhile changes to raise efficiency. So far all we know are the HPC/Fermi capabilities which are hardly any serious indication for the former.
rendezvous
Regular
I wouldn't say NO way, sure, only 5 of the top 500 supercomputers runs windows.
http://www.top500.org/system/details/9787
And heck, given from what I've seen from nexus, it should be worth developing CUDA kernels and C/C++ code on windows. Both are very portable.
Yeah, NVIDIA just spent resources on something irrelevant...Did you by any chance follow GTC at all ? I suggest you listen to Tech-Reports podcast from that time. Scott describes what can only be seen as excitement and interest in what Fermi is (and its potential) and the tools developed for its use, which btw were mostly requests from people working in that market.
Yeah, those are still a mystery. But didn't Rys mention that NVIDIA would talk about Fermi's unknown bits this month ?
Hopefully that's true and we can finally get an idea if the speculation about Fermi's graphics bits, was up to par.
Considering he brings nothing but bad news for Nvidia and your personal slant, I'm not surprised why you'd choose "someone" elses word.
Alternatives does not need to be a better one, for the nth time.
Alternatives does not need to be a better one, for the nth time.
Yes, obviously it makes perfect sense that they would change their minds a month after the announcement because Fermi is too good. Perfect sense.
Silent_Buddha
Legend
Is it just me or does all this furor over proposed paper specs remind one of...
The Hype preceding the release (or non-release) of Nv40, Rampage, Parhelia, R600, the chip by Bitboy Oy that I can't remember the name of, Rendition V3300, etc...
Not saying it's going to follow in the illustrious footsteps of those. But those are the last times I think I've seen so much praise and hope for performance placed upon either paper or rumored specs.
The level of hype for paper specs is astronomical for Fermi.
I fully expect it'll be a good chip, but I think Nvidia would have been better served to have had working silicon seeded to potential customers before even leaking information without NDAs.
Amazing...absolutely amazing...
Regards,
SB
The Hype preceding the release (or non-release) of Nv40, Rampage, Parhelia, R600, the chip by Bitboy Oy that I can't remember the name of, Rendition V3300, etc...
Not saying it's going to follow in the illustrious footsteps of those. But those are the last times I think I've seen so much praise and hope for performance placed upon either paper or rumored specs.
The level of hype for paper specs is astronomical for Fermi.
I fully expect it'll be a good chip, but I think Nvidia would have been better served to have had working silicon seeded to potential customers before even leaking information without NDAs.
Amazing...absolutely amazing...
Regards,
SB
Possibly be but obviously there is line of thinking here that alternatives need to be better or bust!
green.pixel
Veteran
How?
Except for their promises of GF100 being a lot more modular design than previous generations.
Maybe but what are the lead times for building a supercomputer? IIRC they're aiming at 2012.
Except we don't have any specs beyond shader count and bus width. Much of the hype around Fermi is due to it being something new - there still isn't any useful info on its graphics capabilities. And of course whenever something is this delayed the hype grows both from those who want it to succeed and those who want it to fail spectactularly
I can't believe these people who wanted Fermi to fail.
The failure of Fermi will be a disaster for all humankind. Think about the prolonged discovery of cure for cancer and other great scientific achievements to advance humanity. And all they care is tesselation...
Hmm.
I'm not sure. I'm more pro ati than nv30 (the geforce sdr and nv30 kinda hurt me in the pocket) But amd is in dire straits right now and if the fermi is not up to snuff it will allow amd to retain better postion in the market and help stem the losses from the cpu side until hopefully they get more competetive. As Much as I don't want nvidia to go out of busniess its doubly so for amd because they are competition to both nvidia and intel.
Of course considering the size diffrence from the radeons to the fermi it may not really matter in the end as amd will most likely command the price to performance segments of the market.
I can't believe these people who wanted Fermi to fail.
half yoking...how about Nvidia?
Link: http://www.nvidia.com/object/product_tesla_C2050_C2070_us.html
Groo The Wanderer
Regular
Or they could be rethinking the entire project from the ground up. I haven't heard what is happening from here, but I would expect a few more answers this week, unless everyone is on vacation.
-Charlie
