NVIDIA's Fermi: Architected for Tesla, 3 Billion Transistors in 2010
by Anand Lal Shimpi on September 30, 2009 12:00 AM EST- Posted in
- GPUs
Architecting Fermi: More Than 2x GT200
NVIDIA keeps referring to Fermi as a brand new architecture, while calling GT200 (and RV870) bigger versions of their predecessors with a few added features. Marginalizing the efforts required to build any multi-billion transistor chip is just silly, to an extent all of these GPUs have been significantly redesigned.
At a high level, Fermi doesn't look much different than a bigger GT200. NVIDIA is committed to its scalar architecture for the foreseeable future. In fact, its one op per clock per core philosophy comes from a basic desire to execute single threaded programs as quickly as possible. Remember, these are compute and graphics chips. NVIDIA sees no benefit in building a 16-wide or 5-wide core as the basis of its architectures, although we may see a bit more flexibility at the core level in the future.
Despite the similarities, large parts of the architecture have evolved. The redesign happened at low as the core level. NVIDIA used to call these SPs (Streaming Processors), now they call them CUDA Cores, I’m going to call them cores.
All of the processing done at the core level is now to IEEE spec. That’s IEEE-754 2008 for floating point math (same as RV870/5870) and full 32-bit for integers. In the past 32-bit integer multiplies had to be emulated, the hardware could only do 24-bit integer muls. That silliness is now gone. Fused Multiply Add is also included. The goal was to avoid doing any cheesy tricks to implement math. Everything should be industry standards compliant and give you the results that you’d expect.
Double precision floating point (FP64) performance is improved tremendously. Peak 64-bit FP execution rate is now 1/2 of 32-bit FP, it used to be 1/8 (AMD's is 1/5). Wow.
NVIDIA isn’t disclosing clock speeds yet, so we don’t know exactly what that rate is yet.
In G80 and GT200 NVIDIA grouped eight cores into what it called an SM. With Fermi, you get 32 cores per SM.
The high end single-GPU Fermi configuration will have 16 SMs. That’s fewer SMs than GT200, but more cores. 512 to be exact. Fermi has more than twice the core count of the GeForce GTX 285.
Fermi | GT200 | G80 | |
Cores | 512 | 240 | 128 |
Memory Interface | 384-bit GDDR5 | 512-bit GDDR3 | 384-bit GDDR3 |
In addition to the cores, each SM has a Special Function Unit (SFU) used for transcendental math and interpolation. In GT200 this SFU had two pipelines, in Fermi it has four. While NVIDIA increased general math horsepower by 4x per SM, SFU resources only doubled.
The infamous missing MUL has been pulled out of the SFU, we shouldn’t have to quote peak single and dual-issue arithmetic rates any longer for NVIDIA GPUs.
NVIDIA organizes these SMs into TPCs, but the exact hierarchy isn’t being disclosed today. With the launch's Tesla focus we also don't know specific on ROPs, texture filtering or anything else related to 3D graphics. Boo.
A Real Cache Hierarchy
Each SM in GT200 had 16KB of shared memory that could be used by all of the cores. This wasn’t a cache, but rather software managed memory. The application would have to knowingly move data in and out of it. The benefit here is predictability, you always know if something is in shared memory because you put it there. The downside is it doesn’t work so well if the application isn’t very predictable.
Branch heavy applications and many of the general purpose compute applications that NVIDIA is going after need a real cache. So with Fermi at 40nm, NVIDIA gave them a real cache.
Attached to each SM is 64KB of configurable memory. It can be partitioned as 16KB/48KB or 48KB/16KB; one partition is shared memory, the other partition is an L1 cache. The 16KB minimum partition means that applications written for GT200 that require 16KB of shared memory will still work just fine on Fermi. If your app prefers shared memory, it gets 3x the space in Fermi. If your application could really benefit from a cache, Fermi now delivers that as well. GT200 did have an L1 texture cache (one per TPC), but the cache was mostly useless when the GPU ran in compute mode.
The entire chip shares a 768KB L2 cache. The result is a reduced penalty for doing an atomic memory op, Fermi is 5 - 20x faster here than GT200.
415 Comments
View All Comments
Inkie - Saturday, October 3, 2009 - link
Not that I really want to support SD here, but there was working silicon there. It's kind of weird that many sites fail to mention this. Instead, they focus on the mockup.SiliconDoc - Thursday, October 1, 2009 - link
Go read a few articles on how a card is developed, and you'll have the timeline, you red rooster retard.I mean really, I'm talking to ignoramussed spitting cockled mooks.
Please, the articles are right here on your red fan website, so go have a read since it's so important to you how people act when your idiotic speculation is easily and absolutely 100% incorrect, and it's PROVEABLE, the facts are already IN.
gx80050 - Friday, October 2, 2009 - link
You're a fucking friendless loser who should have died on 9/11. Fucking cuntmonomer - Friday, October 2, 2009 - link
In reply to your original link, here's a retraction, of sorts:http://www.fudzilla.com/content/view/15798/1/">http://www.fudzilla.com/content/view/15798/1/
The card Nvidia showed everyone, and said was Fermi is in fact a mock-up. Oh well.
silverblue - Thursday, October 1, 2009 - link
What facts? What framerates can it manage in Crysis? What scores in 3DMark? How good it is at F@H?Link us, so we can all be shown the errors of our ways. It's obvious that GT300 has been benchmarked, or at least, it's only obvious to you simply because the rest of us are on a different planet.
You call people idiots, and then when they reply in a sensible manner, you conveniently forget all that and call them biased (along with multiple variations on the red rooster theme). You're like a scratched vinyl record and it's about time you got off this site if you hate its oh-so-anti-nVidia stance that doesn't actually exist except in your head.
Prove us wrong! Please! I want to see those GT300 benchmarks! Evidence that Anandtech are so far up AMD's rear end that nothing else is worth reporting on fairly!
Zool - Thursday, October 1, 2009 - link
GTX285 had 32 ROPs and 80 TMUs for aorund the same bandwith like 5870 with same 32 ROPs and 80 TMUs. Dont be stupid. GTX will surely need more ROPs and TMUs if they want to keep up with graphic even with the GPGPU bloat.Totally - Wednesday, September 30, 2009 - link
it's 225GB/s not 230.4/s230400/1024 = 225
I'm afraid your bad at math.
Lightnix - Thursday, October 1, 2009 - link
Nope, just really bad at remembering that those prefixes mean 1024 at like 1 in the morning.Lonyo - Wednesday, September 30, 2009 - link
You assume that they will use GDDR5 clocked at the same speed as ATI.They could use higher clocked GDDR5 (meaning even more bandwidth), or lower clocked GDDR5 (meaning less bandwidth).
There's no bandwidth comparison because 1) it's meaningless and 2) it's impossible to make an absolute comparison.
NV will have 50% more bandwidth if the speed of the RAM is the same, but it doesn't have to be the same, it could be higher, or lower, so you can't say what absolute numbers NV will have.
I could make a graph showing equal bandwidth between the two cards even though NV has a bigger bus, or I could make one showing NV having two times the bandwidth despite only a 50% bigger bus.
Both could be valid, but both would be speculative.
Calin - Thursday, October 1, 2009 - link
Also, there could be a chance that the Fermi chip doesn't need/use much more bandwidth than the GT200. Available bandwidth does not performance make