• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Fermi (Nvidia Next Gen) GCPU Architecture: Thread of promises, waiting and 2010

From [H]ard|OCP

NVIDIA had a big get-together in Hong Kong last week where a lot of the NVIDIA upper brass was on hand. It is being rumored that the first "Fermi build kits" were supposed to be prepped to go out to AIBs but NVIDIA missed the mark on having its Fermi tech ready to pass off to builders. We also heard the "Gemini" name being batted around as well and this seems to be connected to a dual GPU Fermi card. We still feel as though NVIDIA is aiming for getting Fermi shipping by the last week of February, but still think it will miss this mark. HardOCP's best guess right now is late March with any quantity to speak and we are still giving NVIDIA about even odds of marking March. April 2010 is looking more reasonable to us at this time.
Click to expand...
Q2 2010.
 
Culex said:
Yuck, this is very disappointing. I wish Nvidia had a similar product out right now, because ATI is RAISING PRICES on it's 5XXX line because of zero competition.
Click to expand...

While that's true (I read ATI raised the price on the 5850 by $20), it's also retailers gouging due to low-supply and high-demand raising it an additional $30.
 
Fuck... I was hoping for something exciting, but it is more Nvidia bashing.
 
More Watts on the fire.

http://www.semiaccurate.com/2009/12/21/nvidia-castrates-fermi-448sps/


IT LOOKS LIKE we were right about Fermi being too big, too hot, and too late, Nvidia just castrated it to 448SPs. Even at that, it is a 225 Watt part, slipping into the future.

Update: Nvidia has contacted us and declined to respond.

The main point is from an Nvidia PDF first found here. On page 6, there are some interesting specs, 448 stream processors (SPs), not 512, 1.40GHz, slower than the G200's 1.476GHz, and the big 6GB GDDR5 variant is delayed until 2H 2010. To be charitable, the last one isn't Nvidia's fault, it needs 64x32GDDR5 to make it work, and that isn't coming until 2H 2010 now.


Fermi - Now less than promised


The rest however is Nvidia's fault. It designed a chip that was more or less unmanufacturable, and we have been saying as much for more than six months now. It is big, 530mm^2 at the minimum, likely 10+mm^2 more, and way too hot. The 448 SP version is listed at 225W TDP, 190W 'typical', and that is with 14 of the 16 shader clusters clusters fused off. With them on, it would likely be well above 250W, far too hot for a single GPU card.

Then there is the whole problem of 512 SPs promised versus 448 SPs delivered. Fermi is arranged as 16 clusters of 32 shaders, and given that it is turning off 64 shaders, it looks like the minimum granularity it can fuse off is a single cluster of 32. This means it is having problems getting less than two unrecoverable errors per die, not a good sign.

Before you wonder if Nvidia is doing this on purpose, the Fermi based Tesla boards cost $2,499 and $3,999 for the small and large ones respectively. If there was ever a case to cherry pick the good parts, it is for low volume halo parts that sell for about 10 times what a gamer card goes for. Nvidia can not get enough yields for the cream of the crop, so how do you think the mainstream will fare? Look for a green spray-painted X800XTX-PE to not grace shelves before this is all said and done.

The architecture is broken, badly designed, and badly thought out. Nvidia does not understand the basics of modern semiconductor design, and is architecting its chips based on ego, not science. The era of massive GPUs is long over, but Nvidia (mis-)management doesn't seem to want to move their egos out of the way and do the right thing. Now the company is left with a flagship part it can't make.

If Nvidia has a saving grace, it is the upcoming 28nm process node from TSMC, but that is for all intents a 2011 process for any sort of volume. Given that ATI tends to beat Nvidia to any new process node, and does it better, that makes it likely that this "light at the end of the tunnel" will be ATI's Northern Islands chips running over it, rather than anything that fixes Nvidia's problems.

Unless Nvidia has a radical new part waiting in the wings for Q2, it really doesn't have much of a chance to win much of anything. Fermi based cards may beat Cypress by a little, but at a hugely higher TDP and a much higher cost. Architecturally speaking, this generation is basically lost, press stunts aside.

If you are waiting to see the fallout of such architectural snafus, look no further. 448 SPs at 225W TDP, massive yield problems from day 1, and an end result that will be barely manufacturable. Look for more backpedaling soon, count on it.S|A
Click to expand...
 
Rabid Wolverine said:
More Watts on the fire.

http://www.semiaccurate.com/2009/12/21/nvidia-castrates-fermi-448sps/


IT LOOKS LIKE we were right about Fermi being too big, too hot, and too late, Nvidia just castrated it to 448SPs. Even at that, it is a 225 Watt part, slipping into the future.

Update: Nvidia has contacted us and declined to respond.

The main point is from an Nvidia PDF first found here. On page 6, there are some interesting specs, 448 stream processors (SPs), not 512, 1.40GHz, slower than the G200's 1.476GHz, and the big 6GB GDDR5 variant is delayed until 2H 2010. To be charitable, the last one isn't Nvidia's fault, it needs 64x32GDDR5 to make it work, and that isn't coming until 2H 2010 now.
Click to expand...
I also heard it eats children.

Man, that reads more like a troll than a serious report.
 
Card is still gonna be a beast, but the cost could still be way to much and the delays don't help it one bit. FX it is not, but this is the biggest dent in the nvidia armor in a long while.

I'm more worried that DX11 will get tagged as crap like DX10 cause you know how bad pr launches can carry through a life cycle even if the product gets good. Just look at Windows Vista.
 
29zpq3a.jpg


http://www.youtube.com/watch?v=gkI-ThRTrPY

its so slow without a hardware tesselato..... oh wait
 
bee said:
yus
Click to expand...

Shouldn't it be a 3XX name?

Anyway, good to see it actually working. Maybe 1Q 2010 will be an actual launch with actual board availability but all the rumors indicate otherwise.

Also, what is with the GT3XXM line of mobile graphics cards? Are they rebranded 2XX series or something new?
 
That's a nice case and watercooling system. It's so neat.
bee said:
its so slow without a hardware tesselato..... oh wait
Click to expand...
All he said was it was running full speed. Did you expect it to be a slideshow? We'll see how it does in games when the cpu is having to deal with all the aspects of a game engine but logic says adding tessellation workload to a general purpose cpu is going to be a bigger performance hit than hardware tessellation. Just like how I can't run physx in Mirror's Edge on my 4850 without a massive performance hit.
 
Blackface said:
Hopefully this is released soon, so ATI can do some price cuts.

Now AMD needs to get it's act together with the CPU's.
Click to expand...

On this note, AMD said at CES that they've shipped their 2 millionth DX 11 card.
 
Looking good, but the big questions that remain are price and availability. I don't think it's going to be too exciting if the MSRP is ~$600 or if a lack of stock drives it up to that kind of price.

And of course real benchmarks. But once these things are all transformed from ?s to reality, there might be a lot of 5850 + 5870s on the market (which may be a good thing in getting prices down further)!
 
What I want to know is how much VRAM Fermi will have? It must have at least 1.5 GB to be interesting to me. I didn't get a 5870 or 5970 from ATI because they still only have 1 GB per GPU and so I wasn't really compelled to upgrade.

Now I am hoping Fermi launches with 1.5-2 GB of VRAM prompting ATI to release a 2 GB version of the 5870 or something like a 5990 with 2x2 GB. Then pricing and benchmarks will decide whether I go with ATI or nVidia.

I can wait for 2GB before jumping on DX11. Got plenty of games that don't DX11 still to be played.
 
DennisK4 said:
What I want to know is how much VRAM Fermi will have? It must have at least 1.5 GB to be interesting to me. I didn't get a 5870 or 5970 from ATI because they still only have 1 GB per GPU and so I wasn't really compelled to upgrade.

Now I am hoping Fermi launches with 1.5-2 GB of VRAM prompting ATI to release a 2 GB version of the 5870 or something like a 5990 with 2x2 GB. Then pricing and benchmarks will decide whether I go with ATI or nVidia.

I can wait for 2GB before jumping on DX11. Got plenty of games that don't DX11 still to be played.
Click to expand...

its gotta be a multiple of the bus and i think the bus is 384bit so its probably 1536mb
 
DennisK4 said:
What I want to know is how much VRAM Fermi will have? It must have at least 1.5 GB to be interesting to me. I didn't get a 5870 or 5970 from ATI because they still only have 1 GB per GPU and so I wasn't really compelled to upgrade.

Now I am hoping Fermi launches with 1.5-2 GB of VRAM prompting ATI to release a 2 GB version of the 5870 or something like a 5990 with 2x2 GB. Then pricing and benchmarks will decide whether I go with ATI or nVidia.

I can wait for 2GB before jumping on DX11. Got plenty of games that don't DX11 still to be played.
Click to expand...

Benches of 1GB of VRAM vs 2GB made no difference except for one or two games afaik, unless you're running 1600p, which maybe you are?
 
Minsc said:
Benches of 1GB of VRAM vs 2GB made no difference except for one or two games afaik, unless you're running 1600p, which maybe you are?
Click to expand...
Yes, I am. And I like High-Res textures, Supersampling AA etc.

I want it for games like ArmA 2, GTA IV and Oblivion and other heavily modded games.
 
Minsc said:
Benches of 1GB of VRAM vs 2GB made no difference except for one or two games afaik, unless you're running 1600p, which maybe you are?
Click to expand...

It's not about benches. It's about being able to use the highest resolution textures in future games.
 
Pimpbaa said:
It's not about benches. It's about being able to use the highest resolution textures in future games.
Click to expand...

I doubt textures will get much more resolution than they are now at the ultra settings. I mean Oblivions 4K textures from three years ago still aren't really topped afaik.
 
bee said:
http://ispss.istreamplanet.com/nvidia/

tegra looks FAST, showed a tablet playing unreal engine 3

whatever mobile device that has one of them in it is gonna be pretty damn good graphically
Click to expand...

Oh sweet so that's what the other mobile UE3 platform is going to be. UE3 on Nintendo's next handheld platform!? Awesome! :D

Tegra powered DS2 is probably my most anticipated hardware launch at this point, can't wait to see what Nintendo's developers can do with competitve modern hardware.
 
xemumanic said:
I thought the general consensus was they were using CUDA cores to do tessellation.
Click to expand...
No, that approach would mean its done in software. ATI had a hardware based tessellator since last 4-5 generations, even Xenos (X360) has one.

Btw, Fermi tech demo of PhysX, Tessellation, DX11: http://www.youtube.com/watch?v=6RdIrY6NYrM

Folks at B3D are pointing out the slide show when the wireframes are shown.
 
irfan said:
Folks at B3D are pointing out the slide show when the wireframes are shown.
Click to expand...
Wireframes are traditionally slow since they're only accelerated in Quadro line drivers. Doesn't mean a thing.
 
dr_rus said:
Wireframes are traditionally slow since they're only accelerated in Quadro line drivers. Doesn't mean a thing.
Click to expand...
Funny thing I didnt see slideshows on Cypress. Again might be something or nothing. We just have no real data on Fermi yet.
 
irfan said:
http://www.anandtech.com/tradeshows/showdoc.aspx?i=3719&p=3

Question is what price? Nvidia's PR group (fudzilla) is mouthing that Fermi is the hottest GPU ever from Nvidia, given their problems with bumpgate and GTX280 failures, not sure that news is comforting.
Click to expand...

Wonder what the chances of it coming out for a MSRP of $399.99 being faster than the 5870? That'd really stir things up!

I'd be a little disappointed if it launches in the $499.99-$599.99 range.
 
Minsc said:
Wonder what the chances of it coming out for a MSRP of $399.99 being faster than the 5870? That'd really stir things up!

I'd be a little disappointed if it launches in the $499.99-$599.99 range.
Click to expand...


It would be surprising.. but given that Nvidia has NEVER launched a product that delivered on both price and performance, I wouldn't get my hopes up. I'm expecting Fermi to cost at least $100 more than ATI's current crop of cards.
 
You must log in or register to reply here.

Similar threads

What's gonna be the Series S of next gen?
2
59
800
Exede replied
  • Poll
Will PS6/Xbox next finally offer true next-gen graphics?
2 3 4 5 6
262
3K
RayHell replied
Intel and AMD next gen CPU delay to 2027
2
73
2K
MCH2024 replied
Discussing Multi Frame Gen(nvidia new game support)
3
184
LiquidMetal14 replied
NVIDIA and the Death of Art Direction
Opinion Trailer 
43
501
ResurrectedContrarian replied
Top Bottom