| Page 1 of 28 |
Veki
Posts: 381
Location: Croatia
|
 Posted: Fri, 2nd Oct 2009 11:24 Post subject: nVidia Fermi - Next Generation CUDA Architecture |
|
 |
| Quote: | The next generation CUDA architecture, code named “Fermi”, is the most advanced GPU computing architecture ever built. With over three billion transistors and featuring up to 512 CUDA cores, Fermi delivers supercomputing features and performance at 1/10th the cost and 1/20th the power of traditional CPU-only servers.
Fermi makes GPU and CPU co-processing pervasive by addressing the full-spectrum of computing applications. Designed for C++ and available with a Visual Studio development environment, it makes parallel programming easier and accelerates performance on a wider array of applications than ever before – including dramatic performance acceleration in ray tracing, physics, finite element analysis, high-precision scientific computing, sparse linear algebra, sorting, and search algorithms.
|
http://www.nvidia.com/object/fermi_architecture.html
White paper : LINK
Specs :- 512 CUDA cores
- MIMD-similar functions
- ECC support
- 1 MB L1 cache
- 768 KB L2 cache
- 384 bit memory bus
- direct hardware access [HAL] for CUDA 3.0, DirectX 11, OpenGL 3.1 and OpenCL
- direct hardware access [HAL] for C (CUDA), C++ and Fortran
- 3 billion transistors
A brave move by nVidia. This new core looks very promising.
Beware of he who would deny you access to information, for in his heart he dreams himself your master.
Commisssioner Pravin Lal
"U.N. Declaration of Rights"
|
|
| Back to top |
|
|
|
|
| Posted: Fri, 2nd Oct 2009 11:59 Post subject: |
|
|
Especially if you're a scientist 
"Quantum mechanics is actually, contrary to it's reputation, unbeliveably simple, once you take the physics out."
Scott Aaronson | chiv wrote: | | thats true you know. newton didnt discover gravity. the apple told him about it, and then he killed it. the core was never found. |
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
|
|
| Posted: Fri, 2nd Oct 2009 22:01 Post subject: |
|
|
can you say realtime raytracing?
about time
i wonder if it's compatible with the intel raytracing tech
|
|
| Back to top |
|
|
Frant
King's Bounty
Posts: 24679
Location: Your Mom
|
| Posted: Fri, 2nd Oct 2009 22:19 Post subject: |
|
|
Yay, nvidia goes for GPU-computing instead of gfx-horsepower. I wonder if this has got something to do with the fact that they're without an x86 license, or if it's just because they have difficulties keeping up with ATI's GPU-designs. Will be interesting to see if they'll do pseudo-emulation of graphics work (like DX11 etc.) through Cuda or if they've actually added all the hardware necessary. I mean, that chip is a dream for scientific applications and encoding, but for gaming???? I don't see it. Gamers don't need supercomputing, they need horsepower to push as many pixels as possible while making them look good at the same time.
Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn!
"The sky was the color of a TV tuned to a dead station" - Neuromancer
|
|
| Back to top |
|
|
Neon
VIP Member
Posts: 18935
Location: Poland
|
| Posted: Fri, 2nd Oct 2009 22:27 Post subject: |
|
|
|
|
|
| Back to top |
|
|
JBeckman
VIP Member
Posts: 35138
Location: Sweden
|
| Posted: Thu, 7th Jan 2010 11:09 Post subject: |
|
|
|
|
|
| Back to top |
|
|
Slizza
Posts: 2345
Location: Bulgaria
|
| Posted: Thu, 7th Jan 2010 11:28 Post subject: |
|
|
Hmmm rumor has it the new nVIDIA chip is 36% faster than a HD5870 after talk from CES 2010 where they had it on show.
I would link but i'm too lazy to go googleing.
Corsair 750D :: 750W DPS-G:: Asus x370 PRO :: R7 1800X ::16gb DDR4 :: GTX 1070::525gb SSD::Coolermaster 240MM AIO::
|
|
| Back to top |
|
|
|
|
| Posted: Thu, 7th Jan 2010 12:11 Post subject: |
|
|
| Slizza wrote: | Hmmm rumor has it the new nVIDIA chip is 36% faster than a HD5870 after talk from CES 2010 where they had it on show.
I would link but i'm too lazy to go googleing. |
the question is, in which benchmark? and then there's the synthetic bench != real life thing. aside from the fact that the fermi will have UP TO 512 cuda cores. who knows, maybe even the flagship card at the release won't have the full 512.
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Thu, 7th Jan 2010 12:55 Post subject: |
|
|
That was from NVIDIA in-house synthetic benchmarking mostly, although they claim 150 FPS or something stupid like that in Dirt 2.
We won't know if Fermi is any good until it comes out, couple more months 
|
|
| Back to top |
|
|
Slizza
Posts: 2345
Location: Bulgaria
|
| Posted: Thu, 7th Jan 2010 15:33 Post subject: |
|
|
They were running that Unigine Heaven benchmark so it may have been referring to that.
Or it may have been a overall statistic.
Or it may have been made up by a industry troll.
Corsair 750D :: 750W DPS-G:: Asus x370 PRO :: R7 1800X ::16gb DDR4 :: GTX 1070::525gb SSD::Coolermaster 240MM AIO::
|
|
| Back to top |
|
|
JBeckman
VIP Member
Posts: 35138
Location: Sweden
|
| Posted: Fri, 8th Jan 2010 16:01 Post subject: |
|
|
|
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Sat, 9th Jan 2010 12:53 Post subject: |
|
|
So far, NVIDIA is losing points big time, they need to start getting some real world numbers out there.
- 3 displays requires SLI? Sorry NVIDIA, but no.
- There's gonna be a Fermi-certified label for new computer cases and GPU coolers, to indicate that they got sufficient cooling for the 300W-using card. What the hell at that 
- Rumour has it that the cards on display there, match the 5870 in performance, but aren't as fast as 5970. To do that, they'd have to increase clocks go beyond 300W (which is the max of the official PCI-E 2.0 spec).
Last 2 points don't come from NVIDIA, but from Thermaltake and Coolermaster representatives. The certification crap sounds very typical, so I'm not surprised at that. If this card really does do 300W under load, then they'd better be wrong about the performance. At 300W I'd expect it to surpass the 5970, which does 210ish under load.
Not good NVIDIA, not good at all so far :E
|
|
| Back to top |
|
|
JBeckman
VIP Member
Posts: 35138
Location: Sweden
|
| Posted: Mon, 18th Jan 2010 17:07 Post subject: |
|
|
|
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Mon, 18th Jan 2010 19:17 Post subject: |
|
|
|
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Jan 2010 08:24 Post subject: |
|
|
|
|
|
| Back to top |
|
|
Sin317
Banned
Posts: 24321
Location: Geneva
|
| Posted: Tue, 19th Jan 2010 08:42 Post subject: |
|
|
|
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Tue, 19th Jan 2010 09:49 Post subject: |
|
|
| Sin317 wrote: | you dont really GET benchmarks and tests, do ya ?
by simply looking at the graph provided there, its clear that the difference won't change a bit ....
ah sorry, you're an ati fanboy i guess lol ... nm ... move along ... |
You've been calling people fanboys, and accusing "ATI fanboys" of talking shit about NVIDIA, but here you are doing exactly that towards ATI.
Moreover, you could not be more wrong.
60 seconds of the Heaven benchmarks means jackshit, since what they did includes the 25 second scene with heavy tessellation. Before that, AA and AF performance are heavily tested, and after that the shaders and texturing gets tested.
So yes, well fucking done NVIDIA, you are doing better than ATI on a piece of technology they were the first ones to work with; hell, they were the ones who pushed it into DX11. Now please show us how you do with all the other (old) techniques that will still be used WAY more than Tessellation in the coming year?
Stop acting like a smartass and calling people fanboys when you're clearly one yourself and don't know what the hell you're talking about.
And in before your reply: No, I'm not an ATI fanboy, I've had more NVIDIA cards than ATI cards, and that's still without counting the RMA cards I had to get.
The only thing you can accuse me of is of being a 3DFX fanboy, Voodoo 1 rocked the world 
|
|
| Back to top |
|
|
LeoNatan
☢ NFOHump Despot ☢
Posts: 73361
Location: Ramat HaSharon, Israel 🇮🇱
|
| Posted: Tue, 19th Jan 2010 10:19 Post subject: |
|
|
BTW, that benchmark is nothing impressive really. Fermi has a dedicated tessellation processor and Unigine makes a heavy use of tessellation.
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Jan 2010 10:30 Post subject: |
|
|
I'm an eyecandy fanboy phuck the politics for phuck sakes!
|
|
| Back to top |
|
|
Slizza
Posts: 2345
Location: Bulgaria
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Jan 2010 10:38 Post subject: |
|
|
woooooooooot! fucking epic! i just heard fermi is going to come bundled with a free copy of!
 Spoiler: | duke nuken forever |
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Jan 2010 10:43 Post subject: |
|
|
|
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
|
| Back to top |
|
|
Slizza
Posts: 2345
Location: Bulgaria
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Jan 2010 14:39 Post subject: |
|
|
YAY! So Fermi is marginally faster than the 5870! WEWT! You, too, can have all that power for the low, low price of $899!!!!!!11
Fuck nVidia. I can't believe some of you are actually falling for their bullshit. Wait, yes I can. nVidia fanboys who have been bawing their eyes out about ATi kicking the shit out of their beloved cards, so now you think you have ammo against ATi.
Let's just see when (chortle, I mean IF) the card actually surfaces.
|
|
| Back to top |
|
|
Sin317
Banned
Posts: 24321
Location: Geneva
|
| Posted: Tue, 19th Jan 2010 16:25 Post subject: |
|
|
| Quote: | | So Fermi is marginally faster than the 5870! |
stuff like 1.5x - 3.3x faster is marginal .. right ...
| Quote: | | nVidia fanboys who have been bawing their eyes out about ATi kicking the shit out of their beloved cards |
where did that happen exactly ? Cause i didn't see the 4xxx series kick the shit out of the gtx2xx series, or the 3xxx series kicking the shit out of the 9xxx series, or the 2xxx series kicking the shit out of the 8xxx series ... yes those are the counterparts ...
Now don't even dare to say "yeah but the 5xxx series kicks the shit out of the gtx2xx series "
Well first off, if you call the Fermi results "marginal" than the 5xxx vs 2xx results are at best "barely noticable" in comparison
Plus, its like comparing a 2xx to a hd3xxx ....
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Jan 2010 17:18 Post subject: |
|
|
Meanwhile I run all games at high (1680x1050) on my gtx260.
Like I give two fucks about the ultrapriced watt-consuming shit Nvidia will churn out, or ATI's shitty drv support.
|
|
| Back to top |
|
|
| Page 1 of 28 |
All times are GMT + 1 Hour |
