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.

A Different Sort of Launch A More Efficient Architecture
Comments Locked

415 Comments

View All Comments

  • SiliconDoc - Wednesday, September 30, 2009 - link

    Nice rebuttal to page 2: " Another kind of LAUNCH "
    --
    write it down, nvidia launched today....(according to lunatic lying red roosters)
  • tamalero - Wednesday, September 30, 2009 - link

    weird.. they still said its "coming soon", I dont see any GF300 firm chips.
    when ATI said "we present the 5870" they were already on newegg.com

    Silicon, let's face it, you're the biggest pro-nvidia troll I've ever seen.
  • SiliconDoc - Wednesday, September 30, 2009 - link

    You are also the person that went into a tirade about nvidia not replacing laptop gpu's with the faulty substrate and instead puttig on a heftier fan.
    You waxed on about how much you hate nvidia, and how they harmed the children (you claimed to be a teacher of some sort) then you screeched about nvidia reps, wished violence upon them, and claimed you'd love to show them how to do their jobs correctly.
    ---
    That's YOU tamalero.
    --
    Now it's pretty amazing I tell the simple plain truth, you deny it a week late, lying for ati, have you public hate and rage on this board for nvidia, and yet claim it is I that is a fanboy.
    --
    One Q, has your raging hatred for nvidia receded, or does lying about the 5870 release give you a sense of vengeful pleasure ?
  • tamalero - Friday, October 2, 2009 - link

    what truth?
    you're just inventing random crap your brain somehow imagines in illusions.
    and what the hell are you talking about?
    I never claimed to be a "teacher", wished violence? what the hell are you smoking?
    harmed the children.. jesuchrist... are you on some sort of scientologist brainwashing group ?

  • SiliconDoc - Friday, October 2, 2009 - link

    Since you have lied, I will get the link and your quotes.
  • SiliconDoc - Wednesday, September 30, 2009 - link

    no they wre not already on newegg - listed and greyed out- the first one available in a trickle -and only today have those listed appeared available, before that it was on for a few seconds, card gone - all GREYED OUT again.
    ---
    Sept. 23rd was launch, this is 7 days later.
    They were a WEEK of paper. (no one can fairly count a sickly 1,2 or half dozen trickle)
  • tamalero - Friday, October 2, 2009 - link

    they were grey, because they sold out, note..., there were on amazon and tigerdirect.com as well. I woudlnt be surprised if newwave and other sites had the 5870 as well.
    you're just a person with mental problems who cant really accept anything outside your tiny world.
  • SiliconDoc - Friday, October 2, 2009 - link

    TigerDirect was pre-order, as well as Amazon was reserve - you just haven't got clue one.
  • fikimiki - Friday, October 2, 2009 - link

    In Poland, (it is Europe cause you don't know for sure)
    it is available in shops.
    Also you can grab one from newegg.com
  • bobvodka - Thursday, October 1, 2009 - link

    I woke up on HD5870 launch day.
    I logged onto a website in the UK.
    I ordered an HD5870.
    It shipped the same day.
    I had it the next day and have been enjoying it ever since.

    Looks like a non-paper launch to me.

Log in

Don't have an account? Sign up now