http://vr-zone.com/articles/how-the-gef ... 15786.html
"The result of GF100 was newly found approach for the subsequent Kepler generation. It was decided to do a die-shrink of Fermi dies such as GF100 (GTX 480), GF104 (GTX 460) etc. giving birth to GF110 (GTX 580), GF114 (GTX 560 Ti) and so on and so forth. In that time, AMD executed Evergreen architecture (Radeon HD 5000), Northern Islands (Radeon HD 6000) and now Southern Islands (Radeon HD 7000). However, NVIDIA had a new approach to launching the parts and for the first time, the company was working on three dies at the same time, what was called high-power, high-efficiency and low-power dies, i.e. entry, mainstream and high-end silicon. The high-end silicon is supposed to debut really soon, on a HPC event on May 10th."

This is impressive


Instead of firing employees in 2009, the senior executives took a major salary reduction. Nvidia is a smart company - unlike their competitor who should have fired the upper management instead of their hard-working employees

I don't think nVidia has been really reshaped but it is AMD Trying to figure out what business model to use -- considering the sheer changes and reshaping going on there.

Why would nvidia be asking Theo Valich if AMD had any disabled parts in Tahiti though?! Methinks the author is suffering from a swollen head there, I'm quite sure nvidia has the means and the engineers to tear down a chip and see for themselves .......

They wouldn't. However, back *before* HD 7970 launched, that question would have been on their minds just as it was on mind when i heard the leaked lack-luster performance and said so in my own preview
- they would also know when they got their hands on it, that there were no disabled parts and then it was full steam ahead with their Kepler plans.

And i just found out how AMD plans to combat it and increase their clockspeeds significantly - using VaporX cooling for some of their SKUs. New Radeon launch next week.

They are trying to hold their margins and i don't believe they will be successful

I thought vapor chambers were standard on highend AMD cards since around R600 timeframe?

http://www.sapphiretech.com/VaporX/VaporX_paper.html

Look for the first one on the HD 7000 series next week

i am not sure this guy knows even what he talks about????




the first one in bold is just wayy off. The 580/560ti is not a die shrink of anything

this article is just a bunch of connecting dots and filling in with imagination. There is way more to it. For starters the performance of the 7970 is not why we didnt see the gk100 as the 680. Many more factors have relevance above that. It was well known that TSMC couldnt deliver 28nm wafers in adequate quantities. Making large 600mm^2 dies in quantity was nothing more than a fantasy. This posed a real issue, one that AMD was most grateful to be aware of.

So AMD thinks nvidia is screwed and it will be at least Q3-Q4 before nvidia can compete with their large dies. There is only so many 28nm wafers TSMC could produce and they were being divided up with multiple companies. They debuted the HD7970 at a premium price, higher than what we have seen flagships cost in a long while. Much higher than AMD has ever charged for a gpu, completely abandoning their sweet spot strategy. Banking on the fact that nvidia has no chance of competing with their large die with TSMC.

Meanwhile, nvidia is completely aware of the shortage. The issue was real, and they had to work within the limitations. The gk104 had some rather large shoes to fill and early on nvidia was aware of this. Early on nvidia decided to push the gk104 far. Originally the GTX670ti, they hoped to get as close to matching AMDs flagship as they could. They would at least price it very competitive and compete until they could finally drop the bigK. This would be their best case..... or so they thought.

The GK104 was a very powerful gpu that was very efficient. It was built lean and mean from the start. But once AMD released the 7970 it became clear to nvidia that they wouldnt have to struggle to be second place. They discovered that AMDs flagship was well with reach of the gk104, that it would be relatively easy to beat it while using less power even. Most of the major structure of the design was there. All nvidia had to do was work out the clocks. They already had a 256bit bus in design, so they had to use very high speed ram. Rather quickly nvidia had the 670ti supercharged to become the gtx680. It would completely catch AMD off guard and it was a hard blow to take.

AMD planned as the predator/aggressor for comfortable margins selling their chips without much competition. Once the 680 launched they scrambled and were very reluctant to lower the prices of the 7970. There was hope that a weakness could be exploited which would keep the 7970 at $550. Finally it becomes clear that the 7970 will have to be reduced in price which effects other cards in their line up. AMD is still fighting denial and they plan a ghz 7970 to keep the margins. In the long run, the gk104 is more power efficient at higher clocks and overclocked 680s will easily combat the ghz 7970. But once again amd is banking on nvidia having limited supplies. This would make the ghz 7970 a great alternative and would give people little reason to wait for nvidia restocks. Its a short term solution but really all solutions are. Eventually all of these cards will be overshadowed by the bigK which is several months away. Its limited supply will make for a high high price. Most of the initial dies will have to go to tesla. As TSMC ramps up production, things will level out. Its been really interesting seeing things unfold. In time the 28nm shortage will no longer be a part of strategic cooperate plans. Maybe then things will be back to normal, whatever normal is.......


anyway, this is my collection of information. Apoppin, you can make an article out of it if you like. Its much more to the roots than theo, i assure you.

i thought the 107 was a typo

i know about this and it will be in my very next evaluation .. this week

My sources have clarified everything and i will get much more information at the GTC

Thank-you

its a little different than theo's version. Well maybe a lot different. I know you were asking a lot of the right questions all along. you were either on the right track or doing it for rhetoric. I knew you should have known most of all of this, i just kinda put it in order. After i seen theos half as article i couldnt help but put together words that are more to the roots.

it would be nice to see a more complete truth filled version in an article apoppin. Bring in the behind the scenes working. You could actually use this stuff for its own standing. In a blog or news. That could bring in traffic by itself. But i know your pressed for time so combining it makes sense. Its just that most ppl skim over benchmark reviews and only focus on the charts.

I at least think you should clarify some of this stuff. This angle covers both sides and is much more grounded in the center.

Love to read what you have gathered. Everyone gets to hear stuff from different places and sometimes we get the same information but different angles. The 28nm node has been very different than other shifts. I find it really interesting and a landmark. Nvidia is making a huge change now because of the way 28nm has limited them. This will be the last time i believe all these companies will be stuck with TSMC only. Anyway its been monumental and i think its a very interesting chain of events.