AMD Radeon HD 7970 Review: 28nm And Graphics Core Next, Together As One
by Ryan Smith on December 22, 2011 12:00 AM EST- Posted in
- GPUs
- AMD
- Radeon
- ATI
- Radeon HD 7000
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.
Facebook
Twitter
RSS
292 Comments
View All Comments
RussianSensation - Saturday, January 14, 2012 - link
BF3 is not a 2012 game.......Also, most of us have been gaming on our older cards. Who in the world who has a previous high-end card is going to drop $600 for BF3 alone? No thanks.
SSIV - Saturday, February 18, 2012 - link
Since there's a new driver out for there cards we can now regard these results with a grain of salt. Revise the benchmarks!DaOGGuru - Thursday, March 1, 2012 - link
I don't know why people keep forgeting about the 560ti 2win. Yes I said 2win = 2 560ti processors on one card. It still kills the 7970 numbers in BF3 by 20Fps. and is same price. It also beats the 580 and is cheaper. It's a single card with 50amp min. draw and it will smoke anything except 590 and the 6990...http://www.guru3d.com/article/evga-geforce-gtx-560...
CeriseCogburn - Thursday, March 8, 2012 - link
Oh, right, well this isn't an nvidia card review, so we won't hear from 50 posts about how some CF (would be SLI of course in this case) combo will whip the crap out of it in performance and price...You know ?
That's how it goes...
Usually the articel itself rages on about how some amd CF combo is really so much good and better and blah blah blah.... then the rpice perf, then the results - on and on and on ....
---
The angry ankle biters are swarmed up on the under red dog radeon side...
--
So you made a very good point, I'm just sorry it took 29 pages of reading to get to it, in it's glorious singularity.... you shouldn't strike out in independent thought like that it's dangerous.... not allowed unless the card being reviewed is an nvidia !!!!
DaOGGuru - Thursday, March 1, 2012 - link
oops... forgot to say look at previous post links BF3 rating for the 560ti 2win and compare to this charts 7970 fps. The 2win is pumping out @20 more FPS and is $50.00 - $100.00 cheaper than the 7970... lame.. ATi is still behind Nvidia but proud of it! lol They are just now catching up to Nvidia's tessellations and oh and AFTER they changed to a "cuda core copy" architecture and posting it as big news... Evga's older 560ti 2win still dusts it by 20FPS.. lame.DaOGGuru - Thursday, March 1, 2012 - link
sorry 10FPS not 20.. it's late.DaOGGuru - Thursday, March 1, 2012 - link
I don't get what's the hub-bub about the 7970.. sure it's the fastest single cpu;BUT, for $50.00-$100.00 less you can get the 560Ti 2win (dual cpu) that smokes the 7970 and the 2win PCB does have an SLI bridge and is cabapable of doing SLI to a second card but it's currently locked by Nvidia (see paragraph 3).Also, the 2win draws a min of only 50amps (way less than most sli configurations) 1. has a considerably lower noise dba, 2. runs cooler and with less power than almost all the high end cards and 3. will run 3 montiors in Nvidia 2D and 3D surround off a single card! 4.Will kill the GTX 580 by @33-23% (depending on review) 5. Will beat the 590 in some sample testing for TDP. And finally 6. will kill the 7970 by 10-20FPS in BF3 including by 10FPS in 1920x1200 4AA-16AF Ultra high mode. So, why have people forgotten the 2win? It's a singlecard, multi-GPU, full 3D/2D surround without a second card in SLI, $500.00USD beast !
OH and for those that say you can't SLI with a second 2win.... http://www.guru3d.com/article/evga-geforce-gtx-560... (this review states on conclusion page) > quote " you will have noticed there is a SLI connector on the PCB. Unfortunately you can not add a second card to go for quad-SLI mode. It's not a hardware limitation, yet a limitation set by NVIDIA, the GTX 560 Ti series is only allowed in 2-way SLI mode, which this card already is."
... So actually, the card is cabale 2card SLI but Nvidia for some (gosh aweful reason) won't let the dog off the chain. Probably because it will absolutely kill the need for a GTX580, 570, 560 Ti SLI configuration for ever!
Resources: (pay attention to BF3 FPS and compare to 7970 FPS in this article.)
http://www.anandtech.com/show/5048/evgas-geforce-g...
http://www.guru3d.com/article/evga-geforce-gtx-560...
Peace...
DaOGGuru - Thursday, March 1, 2012 - link
I don't get what's the hub-bub about the 7970.. sure it's the fastest single CPU; BUT, for $50.00-$100.00 less you can get the 560Ti 2win (dualCPU) that smokes the 7970 and the 2win PCB does have an SLI bridge and is capable of doing SLI to a second card but it's currently locked by Nvidia (see paragraph 3).Also, the 2win draws a min of only 50amps (way less than most sli configurations) 1. Has a considerably lower noise DBA, 2. runs cooler and with less power than almost all the high end cards and 3. Will run 3 monitors in Nvidia 2D and 3D surround off a single card! 4.Will kill the GTX 580 by @33-23% (depending on review) 5. Will beat the 590 in some sample testing for TDP. And finally 6. will kill the 7970 by 10-20FPS in BF3 including by 10FPS in 1920x1200 4AA-16AF Ultra high mode. So, why have people forgotten the 2win? It's a single card, multi-GPU, full 3D/2D surround without a second card in SLI, $500.00USD beast !
OH and for those that say you can't SLI with a second 2win.... http://www.guru3d.com/article/evga-geforce-gtx-560... (this review states on conclusion page) > quote " you will have noticed there is a SLI connector on the PCB. Unfortunately you cannot add a second card to go for quad-SLI mode. It's not a hardware limitation, yet a limitation set by NVIDIA, the GTX 560 Ti series is only allowed in 2-way SLI mode, which this card already is."
... So actually, the card is capable 2card SLI but Nvidia for some (gosh awful reason) won't let the dog off the chain. Probably because it will absolutely kill the need for a GTX580, 570, 560 Ti SLI configuration forever!
Resources: (pay attention to BF3 FPS and compare to 7970 FPS in this article.)
http://www.anandtech.com/show/5048/evgas-geforce-g...
http://www.guru3d.com/article/evga-geforce-gtx-560...
Peace...
CeriseCogburn - Thursday, March 8, 2012 - link
Ummm.... I read you, I see your frustration with all the posts - just refer to my one above there - you really should not be dissing the new amd like that - they like are 1st and uhh... nvidia is evil... so no comparisons like that are allowed when the fanboy side content is like 100 to 1....Now next nvidia card review you will notice a hundred posts on how this or that CF beats the nvidia in price perf and overall perf, etc, and it will be memorized and screamed far and wide...
Just like... your point "doesn't count", okay ?
It's best to ignore you GREEN fanboy types... ( yes even if you point out gigantic savings, or rather especially when you do...)
Thanks for waiting till page 30 - a wise choice.
CeriseCogburn - Sunday, March 11, 2012 - link
Southern Islands is a whole generation late. AMD promised us this SI in the last generation 6000 series. Then right before that prior release, they told us they had changed everything and 6000 was not Southern Islands anymore. LOLTalk about late - it's what two years late ?
Maybe it's three years....
In every case here, Nvidia beat them to the core architecture by two years. Now amd is merely late to the party crashing copycats....
That's late son, that's not original, that's not innovative, that's not superior, it's tag a long tu loo little sister style.