• What
    is this?

    You've landed on the AMD Portal on AnandTech. This section is sponsored by AMD. It features a collection of all of our independent AMD content, as well as Tweets & News from AMD directly. AMD will also be running a couple of huge giveaways here so check back for those.

    PRESENTED BY

A Quick Refresher, Cont

Having established what’s bad about VLIW as a compute architecture, let’s discuss what makes a good compute architecture. The most fundamental aspect of compute is that developers want stable and predictable performance, something that VLIW didn’t lend itself to because it was dependency limited. Architectures that can’t work around dependencies will see their performance vary due to those dependencies. Consequently, if you want an architecture with stable performance that’s going to be good for compute workloads then you want an architecture that isn’t impacted by dependencies.

Ultimately dependencies and ILP go hand-in-hand. If you can extract ILP from a workload, then your architecture is by definition bursty. An architecture that can’t extract ILP may not be able to achieve the same level of peak performance, but it will not burst and hence it will be more consistent. This is the guiding principle behind NVIDIA’s Fermi architecture; GF100/GF110 have no ability to extract ILP, and developers love it for that reason.

So with those design goals in mind, let’s talk GCN.

VLIW is a traditional and well proven design for parallel processing. But it is not the only traditional and well proven design for parallel processing. For GCN AMD will be replacing VLIW with what’s fundamentally a Single Instruction Multiple Data (SIMD) vector architecture (note: technically VLIW is a subset of SIMD, but for the purposes of this refresher we’re considering them to be different).


A Single GCN SIMD

At the most fundamental level AMD is still using simple ALUs, just like Cayman before it. In GCN these ALUs are organized into a single SIMD unit, the smallest unit of work for GCN. A SIMD is composed of 16 of these ALUs, along with a 64KB register file for the SIMDs to keep data in.

Above the individual SIMD we have a Compute Unit, the smallest fully independent functional unit. A CU is composed of 4 SIMD units, a hardware scheduler, a branch unit, L1 cache, a local date share, 4 texture units (each with 4 texture fetch load/store units), and a special scalar unit. The scalar unit is responsible for all of the arithmetic operations the simple ALUs can’t do or won’t do efficiently, such as conditional statements (if/then) and transcendental operations.

Because the smallest unit of work is the SIMD and a CU has 4 SIMDs, a CU works on 4 different wavefronts at once. As wavefronts are still 64 operations wide, each cycle a SIMD will complete ¼ of the operations on their respective wavefront, and after 4 cycles the current instruction for the active wavefront is completed.

Cayman by comparison would attempt to execute multiple instructions from the same wavefront in parallel, rather than executing a single instruction from multiple wavefronts. This is where Cayman got bursty – if the instructions were in any way dependent, Cayman would have to let some of its ALUs go idle. GCN on the other hand does not face this issue, because each SIMD handles single instructions from different wavefronts they are in no way attempting to take advantage of ILP, and their performance will be very consistent.


Wavefront Execution Example: SIMD vs. VLIW. Not To Scale - Wavefront Size 16

There are other aspects of GCN that influence its performance – the scalar unit plays a huge part – but in comparison to Cayman, this is the single biggest difference. By not taking advantage of ILP, but instead taking advantage of Thread Level Parallism (TLP) in the form of executing more wavefronts at once, GCN will be able to deliver high compute performance and to do so consistently.

Bringing this all together, to make a complete GPU a number of these GCN CUs will be combined with the rest of the parts we’re accustomed to seeing on a GPU. A frontend is responsible for feeding the GPU, as it contains both the command processors (ACEs) responsible for feeding the CUs and the geometry engines responsible for geometry setup. Meanwhile coming after the CUs will be the ROPs that handle the actual render operations, the L2 cache, the memory controllers, and the various fixed function controllers such as the display controllers, PCIe bus controllers, Universal Video Decoder, and Video Codec Engine.

At the end of the day if AMD has done their homework GCN should significantly improve compute performance relative to VLIW4 while gaming performance should be just as good. Gaming shader operations will execute across the CUs in a much different manner than they did across VLIW, but they should do so at a similar speed. And for games that use compute shaders, they should directly benefit from the compute improvements. It’s by building out a GPU in this manner that AMD can make an architecture that’s significantly better at compute without sacrificing gaming performance, and this is why the resulting GCN architecture is balanced for both compute and graphics.

A Quick Refresher: Graphics Core Next Building Tahiti & the Southern Islands
POST A COMMENT

291 Comments

View All Comments

  • mavere - Thursday, December 22, 2011 - link

    Seriously!

    With the prevalence of practically silent PSUs, efficient tower heatsinks, and large quiet fans, I cannot fathom why the noise floor is 37.9 dB.
    Reply
  • Finally - Thursday, December 22, 2011 - link

    As usual, AT is shooting straight for the brain-dam, I mean, ENTHUSIAST crowd feat. a non-mentioned power supply that should be well around 1000W in order to drive over-priced CPUs as well as quadruple GPU setups.
    If you find that horrendous they will offer you not to read this review, but their upcoming HTPC review where they will employ the same 1000W power supply...
    Reply
  • B3an - Thursday, December 22, 2011 - link

    *face palm*

    1: 1000+ Watt PSU's are normally more quiet if anything as they're better equipped to deal with higher power loads. When a system like this uses nowhere near the PSU's full power the fan often spins at a very low RPM. Some 1000+ PSU's will just shut the fan off completely when a system uses less than 30% of it's power.

    2: It's totally normal for a system to be around 40 dB without including the graphics cards. Two or 3 fans alone normally cause this much noise even if they're large low RPM fans. Then you have noise levels from surroundings which even in a "quiet" room are normally more than 15 dB.

    3: Grow some fucking brain cells kids.
    Reply
  • andymcca - Thursday, December 22, 2011 - link

    1) If you were a quiet computing enthusiast, you would know that the statement
    "1000+ Watt PSU's are normally more quiet if anything"
    is patently false. 1000W PSUs are necessarily less efficient at realistic loads (<600W at full load in single GPU systems). This is a trade-off of optimizing for efficiency at high wattages. There is no free lunch in power electronics. Lower efficiency yields more heat yields more noise, all else being equal. And I assure you that a high end silent/quiet PSU is designed for low air flow and uses components at least as high in quality as their higher wattage (non-silent/non-quiet) competitors. Since the PSU is not decribed (a problem which has been brought up many times in the past concerning AT reviews), who knows?

    2) 40dB is fairly loud if you are aiming for quiet operation. Ambient noise in a quiet room can be roughly 20dB (provided there is not a lot of ambient outdoor noise). 40dB is roughly the amplitude of conversation in a quiet room (non-whispered). A computer that hums as loud as I talk is pretty loud! I'm not sure if you opinion is informed by any empirical experience, but for precise comparison of different sources the floor should be at minimum 20dB below the sources in question.

    3) You have no idea what the parent's age or background is, but your comment #3 certainly implies something about your maturity.
    Reply
  • formulav8 - Tuesday, February 21, 2012 - link

    Seriously grow up. Your a nasty mouth as well. Reply
  • piroroadkill - Thursday, December 22, 2011 - link

    Haha, yeah.

    Still, I guess we have to leave that work to SPCR.
    Reply
  • Kjella - Thursday, December 22, 2011 - link

    High-end graphics cards are even noisier, so who cares? A 250W card won't be quiet no matter what. Using an overclocked Intel Core i7 3960X is obviously so the benchmarks won't be CPU limited, not to make a quiet PC. Reply
  • Ryan Smith - Thursday, December 22, 2011 - link

    Our testing methodology only has us inches from the case (an open case I should add), hence the noise from our H100 closed loop radiator makes itself known. In any case these numbers aren't meant to be absolutes, we only use them on a relative basis. Reply
  • MadMan007 - Thursday, December 22, 2011 - link

    [AES chart] on page 7? Reply
  • MadMan007 - Thursday, December 22, 2011 - link

    More stuff missing on page 9:

    [AF filter test image] [download table]
    Reply

Log in

Don't have an account? Sign up now