The RV870 Story: AMD Showing up to the Fight
by Anand Lal Shimpi on February 14, 2010 12:00 AM EST- Posted in
- GPUs
The Cost of Jumping to 40nm
This part of the story could almost stand on its own, but it directly impacts much of what happened with Cypress and the entire Evergreen stack that it’s worth talking about here.
By now you’ve most likely heard about issues with TSMC’s 40nm process. While the word is that the issues are finally over, poor yields and a slower than expected ramp lead to Cypress shortages last year and contributed to NVIDIA’s Fermi/GF100 delay. For the next couple of pages I want to talk about the move to 40nm and why it’s been so difficult.
The biggest issue with being a fabless semiconductor is that you have one more vendor to deal with when you’re trying to get out a new product. On top of dealing with memory companies, component manufacturers and folks who have IP you need, you also have to deal with a third party that’s going to actually make your chip. To make matters worse, every year or so, your foundry partner comes to you with a brand new process to use.
The pitch always goes the same way. This new process is usually a lot smaller, can run faster and uses less power. As with any company whose job it is to sell something, your foundry partner wants you to buy its latest and greatest as soon as possible. And as is usually the case in the PC industry, they want you to buy it before it's actually ready.
But have no fear. What normally happens is your foundry company will come to you with a list of design rules and hints. If you follow all of the guidelines, the foundry will guarantee that they can produce your chip and that it will work. In other words, do what we tell you to do, and your chip will yield.
Global Foundries' 2010 - 2011 Manufacturing Roadmap
The problem is that if you follow every last one of these design rules and hints your chip won’t be any faster than it was on the older manufacturing process. Your yield will be about the same but your cost will be higher since you’ll bloat your design taking into account these “hints”.
Generally between process nodes the size of the wafer doesn’t change. We were at 200mm wafers for a while and now modern fabs use 300mm wafers. The transistor size does shrink however, so in theory you could fit more die on a wafer with each process shrink.
The problem is with any new process, the cost per wafer goes up. It’s a new process, most likely more complex, and thus the wafer cost is higher. If the wafer costs are 50% higher, then you need to fit at least 50% more die on each wafer in order to break even with your costs on the old process. In reality you actually need to fit more than 50% die per wafer on the new process because yields usually suck at the start. But if you follow the foundry’s guidelines to guarantee yield, you won’t even be close to breaking even.
The end result is you get zero benefit from moving to the new process. That’s not an option for anyone looking to actually use Moore’s Law to their advantage. Definitely not for a GPU company.
The solution is to have some very smart people in your company that can take these design rules and hints the foundry provides, and figure out which ones can be ignored, and ways to work around the others. This is an area where ATI and NVIDIA differ greatly.
Facebook
Twitter
RSS
132 Comments
View All Comments
tomoyo - Monday, February 15, 2010 - link
Another awesome article about the real situation behind the hardware from you Anand! I was on the USS Hornet and wish I had talked to you, but it was a great time nonetheless. It's interesting the change in their thought process between the RV770 and RV870, I hope they keep the winning streak up for the next refresh cycle (which hopefully will stay on the market bulges).WT - Monday, February 15, 2010 - link
*sigh* ^^^There's always one in the crowd.
Take care in the fact that you are the only person who hasn't enjoyed this read.
MegaManX4 - Monday, February 15, 2010 - link
Reminds me much of the Anglo-Saxon "documantaries", where it is always of tertiary relevance WHAT is actually discussed, but it is always of utmost interest how the responsible person "feels" about what he is just seeing, other than just stating the facts.There seems to be a huge crowd vowing for that kind of journalism, Whatever pleases the canaille.
"Jedem das Seine" or "to each his own" then
MegaManX4 - Monday, February 15, 2010 - link
This was actually the worst article i have ever read at anandtech. I know that you Americans always strive for emotionally .Driven stories, but this outright borders on silly exaggeration."Heroes of our Industry", what a Schmalz.
Also, if one would take the real informations presented in that article, it wouldn't justify even a 2 Page Article, let alone that 11 Page behemoth.
They are engineers, they do their jobs. Nothing more, nothing less.
Greetings from Germany
blowfish - Monday, February 15, 2010 - link
hmm, with an attitude like that you'll never get past middle management!Like most here, I loved this article. Anand obviously has the friendship and respect of some very senior players, and we were treated to some great insights into how things work at AMD ATI.
As the reader, you can choose to read or not read the article, simple as that. Maybe you should up your medication.
MegaManX4 - Monday, February 15, 2010 - link
unreasonable polemicpmonti80 - Monday, February 15, 2010 - link
You are the one being unreasonable. This may not be a "scientifically written" article, but no one is claiming it to be. And that's the reason this article is so interesting.saiga6360 - Thursday, February 18, 2010 - link
Apparently German engineers are just soulless robots. His confusion is understandable.BelardA - Monday, February 15, 2010 - link
I enjoyed this article even more than the RV770. I do recommend that everyone read that one too.Kind of shocking that Nvidia didn't use that info from the RV770 article to learn to NOT make big GPUs like the GTX 2xx. yeah yeah, it takes 2-4 years to design a chip.
I thank ATI (and AMD) for not playing marketing games like Nvidia does... I think they have a bigger marketing department than engineers nowadays. They started with the GF2-MX 400 & GF4-MX cards (which were re-labeled updated GF2 cards that were not up to GF3 standards)... but the latest cluster-muck of Nvidia products is nothing but a mess. 8800 re-badged as a 9800 re-badged into the gts 250. Code-name of NVxx go to G80 to G92 to G100. The GT-1xx products that are actually low-end 9xxx products, same with most G200 & G300. I'm not going to be surprised when the GTX 285 gets renamed into the GTS450 at $200! I've seen people who bought the GTS250 and post on the internet "why isn't my new gts250 much faster than my old 8800GT"... because you bought a faster version of your card and thought it was something new. Wow, 3 years with 3 names for the same product, that is marketing.
ATI does good with the entire 4000 series being DX 10.1 products and 5000s are DX11. (Does anyone really use HD-5xxx?) It doesn't feel like ATI is pulling our chain with their products.
AMD should be learning from ATI, they are getting better with CPUs - 2 years late, but AMD CPUs are now faster than Core2 and compete well against the lower end intel i-confused model CPUs. There is still room for improvement which was recommend to them some time ago, but AMD is just going to come out with a new design for next year. But had AMD tweaked their CPUs a bit for another 10~20% performance, they'd be up there with i7s.
I hope in the next ATI GPU, some form of Physics engine is added to go up against nvidia's PhsyX. But perhaps that'll be part of DX12... but Microsoft no longer supports Games for Windows.
Actually, with more and more games going ONLY to consoles, I don't think the need for high-end gaming cards will be needed anymore in the next few years. If there are no games, who needs a $300 3D Gaming card?
Zink - Monday, February 15, 2010 - link
Would also like to say great article. I can't wait for new distributed computing cores come out optimized for ATI's architectures.