TechPowerup - Nvidia Kepler GK104 PCB Drawings and power connector pics

[tviceman]

I find it ironically interesting that Nvidia has a legitimate opportunity to catch, or surpass, AMD in the perf per mm^2 category and also in perf per watt. I'm not saying they will do both, or either, but the chatter out there that GK104 is shaping up to be pretty amazing.

 

[RussianSensation]

For AMD's "sweet spot" die strategy to catch up to NVIDIA's "as big as it can be made" strategy.

Prior to AMD acquiring ATI, ATI didn't care for "small die strategy."

We got:

9700Pro (best GPU at the time)
9800Pro/9800XT (best at the time)
X800XT PE / X850XT PE (arguably better than 6800U, but can be easily considered at least as good)
X1800XT PE/X1900XT/ X1950XTX, etc. (arguably better than 7800/7900 series)
* ATI made a crapload of $, with a market cap the size of AMD.
** ATI was viewed as a premium brand.

After AMD acquired ATI and decided to focus on the small-die strategy:

2900XT (huge flop)
HD3870 (huge flop)
4870, 5870, 6970 (2nd best, relegated to competing on price).
* AMD's GPU division struggles to make $
** AMD's GPUs are viewed as a price/performance brand - brand value erosion
*** AMD's discrete GPUs continue to lose market share

There is no evidence at all that AMD's 'small die' strategy is working, not financially, not from a strong brand image perspective, not from market share. One might conclude ATI knew how to both make its videocards and manage its GPU brand MUCH better for us gamers. My gut feeling is the main reason AMD's upper-level mgmt is forcing the GPU guys to focus so much on performance/watt is because their APU strategy depends on it. NV doesn't make APUs, so they can make GPUs with as large as they need to. The end result has been NV beating AMD with 8800GTX, GTX280, GTX285, GTX480, GTX580 for 3 consecutive generations....that's got to hurt AMD's brand image in the long-term.

That's why I don't understand why people keep bringing up that 'small die size' is better. If someone wants to focus so much on advantages of smaller die sizes, they should create a separate thread.

 

[Keysplayr]

That's why I don't understand why people keep bringing up that 'small die size' is better.

RS. They went with the small die price performance argument since 4870. It's now etched in their memories as a historically proven fact. So, no point in steering now, unfortunately. Just can't wait til Keps gets here. :thumbsup:

 

[tviceman]

One could argue that AMD's small die strategy is working, at least from a point of view regarding execution. They have introduced the hd3000's, hd4000's, hd5000's, hd6000's, and hd7000's while in that same time Nvidia has executed only two new chip architectures (GT200 and Fermi, with refreshes to both architectures). Then again, even with all the new architectures AMD has released, Nvidia has consistently held the single GPU performance crown.

 

[Grooveriding]

Prior to AMD acquiring ATI, ATI didn't care for "small die strategy."

We got:

9700Pro (best GPU at the time)
9800Pro/9800XT (best at the time)
X800XT PE / X850XT PE (arguably better than 6800U, but can be easily considered at least as good)
X1800XT PE/X1900XT/ X1950XTX, etc. (arguably better than 7800/7900 series)
* ATI made a crapload of $, with a market cap the size of AMD.
** ATI was viewed as a premium brand.

After AMD acquired ATI and decided to focus on the small-die strategy:

2900XT (huge flop)
HD3870 (huge flop)
4870, 5870, 6970 (2nd best, relegated to competing on price).
* AMD's GPU division struggles to make $
** AMD's GPUs are viewed as a price/performance brand - brand value erosion
*** AMD's discrete GPUs continue to lose market share

There is no evidence at all that AMD's 'small die' strategy is working, not financially, not from a strong brand image perspective, not from market share. One might conclude ATI knew how to both make its videocards and manage its GPU brand MUCH better for us gamers. My gut feeling is the main reason AMD's upper-level mgmt is forcing the GPU guys to focus so much on performance/watt is because their APU strategy depends on it. NV doesn't make APUs, so they can make GPUs with as large as they need to. The end result has been NV beating AMD with 8800GTX, GTX280, GTX285, GTX480, GTX580 for 3 consecutive generations....that's got to hurt AMD's brand image in the long-term.

That's why I don't understand why people keep bringing up that 'small die size' is better. If someone wants to focus so much on advantages of smaller die sizes, they should create a separate thread.

I think you're right that abandoning a big die was a bad move. ATI had more street cred up until the 2XXX line. The big die halo card may be a niche purchase, but it certainly has value in how it reflects on your brand. Tons of nvidia fans who can't/won't afford 8800GTX, 480s, 580s but latch on to the brand and sing its praises because someone out there has those cards and they're the best.

I remember the 9800PRO, X1XXX lines and ATI was arguably more popular than nvidia then. I had hope before 6970 hit with the rumours of a larger die that they would go balls out, but they didn't.

I can't see how anyone cares about the die size of their chip, beyond how it translates to performance, if they are aware of such things. Some of the argument you are addressing is likely based in that AMD has had a counter to everything nvidia has had out since 4XXX, just the not single-gpu spot. Again, why it seems putting in the effort to put out a big-die is worth it, even though it accounts for 1-2% of purchases, people will buy your lesser cards just because you have the fastest card - even if they don't own it. It's like the guy who blows $35K on the crappy BMW when he could of got a car just as good for $25K, but hey, it's a beamer.

I am getting a chuckle out of the support for GK104 though. Have seen near universal assent that this card will not be good enough to dethrone the 7970, but people are looking forward to an nvidia card for once to get a card they can afford. Usually is the other way around, or waiting for nvidia to release their mid-range after they put out their $600 beast.

It is a smart move on nvidia's part being late with their flagship. They clearly learned and came up with a better strategy after going through the same issues on 40nm. Had the 460 come out in Jan/2010 and the 480 in a better state in August/2010, there would not be the blight of JHH and his grilling machine, wood screws and all.

 

[Vesku]

I think I've been very clear about why I bring it up. When saying something along the lines of "it took AMD a node to beat NVIDIA top end" the die size is a factor. All NVIDIA would have is marketing, period, if a 350mm2 same node die was meeting or beating their 500mm2 same node die. NVIDIA's own investor statements indicate their will be no special pricing with TSMC (they said they have been on wafer pricing since sometime in the 40nm era), therefore they have to deal with the same engineering vs economics reality as any other TSMC customer.

How can I make this easier to understand: It's valid to not care about the engineering underpinnings when talking about consumer desires (Price, Performance, Heat and Electricity use, Features). It is not valid to ignore engineering underpinnings when discussing NVIDIA and AMD performance on a technical level ("LOL it took them x months and a new node to catch NVIDIA) <- the die size becomes an issue when making such a statement. It should be "LOL it took them x months and a new node for AMD's 350mm2 part to beat NVIDIA's 500mm2 part.

Now if you qualify it in terms of FPS or GPGPU and talk about GPU tasks in terms of things like per mm2, per Watt, etc. I can get behind that.

*I'm using place holder numbers, please do not focus on the technical specs and instead consider my argument as a whole.

>> In relation to this thread, I am really excited for when the GK104 arrives. Because it will be the best showdown between AMD and NVIDIA in a long while. In terms of pricing, power draw, cooling solution, basically everything. This is not the "big as it will" go NVIDIA product, that will deserve it's own discussion when there is more substantial information about it.

Prior to AMD acquiring ATI, ATI didn't care for "small die strategy."

We got:

9700Pro (best GPU at the time)
9800Pro/9800XT (best at the time)
X800XT PE / X850XT PE (arguably better than 6800U, but can be easily considered at least as good)
X1800XT PE/X1900XT/ X1950XTX, etc. (arguably better than 7800/7900 series)
* ATI made a crapload of $, with a market cap the size of AMD.
** ATI was viewed as a premium brand.

After AMD acquired ATI and decided to focus on the small-die strategy:

2900XT (huge flop)
HD3870 (huge flop)
4870, 5870, 6970 (2nd best, relegated to competing on price).
* AMD's GPU division struggles to make $
** AMD's GPUs are viewed as a price/performance brand - brand value erosion
*** AMD's discrete GPUs continue to lose market share

There is no evidence at all that AMD's 'small die' strategy is working, not financially, not from a strong brand image perspective, not from market share. One might conclude ATI knew how to both make its videocards and manage its GPU brand MUCH better for us gamers. My gut feeling is the main reason AMD's upper-level mgmt is forcing the GPU guys to focus so much on performance/watt is because their APU strategy depends on it. NV doesn't make APUs, so they can make GPUs with as large as they need to. The end result has been NV beating AMD with 8800GTX, GTX280, GTX285, GTX480, GTX580 for 3 consecutive generations....that's got to hurt AMD's brand image in the long-term.

That's why I don't understand why people keep bringing up that 'small die size' is better. If someone wants to focus so much on advantages of smaller die sizes, they should create a separate thread.

 

[f1sherman]

I find it almost unbelievable that AMD has not made a better business with
HD 4000, 5000 and 6000.

They have been very competitive, maybe even too competitive against their own good. Brand erosion as RS put it.

An average user that bought 5850 2 years ago, had little incentive to buy anything from AMD down the line.
And there is STILL nothing for him in AMD's basket.

But then again I don't buy for a sec notion of Cypress being under-priced to build base, grab market share and similar nonsense.
If it had been under-priced, re-sellers would quickly correct that.


God help them if they lost perf/mm2. They won't have a single fortress to run to.

 

[mak360]

If bigger die is soo much better, why the sudden turn from nvidia?

The whole world/industries are going smaller in die with more performance at the same power, apart from nvidia (we shall see if there is any change this round) or maybe nvidia is correct and everyone and his dog is wrong.

PS Wait and see what happens at smaller nodes

 

[f1sherman]

I doubt it's sudden turn from nvidia. Its the other way around.

AMD is the one who went from pure gaming/rendering/entertainment chip to HPC territory.

And Nvidia will again have their behemoth, they need it for Teslas/Quadro,
do not doubt for a sec they won't

 

[mak360]

Seriously, who cares if it&#8217;s a mid-rage or bottom? (Die size and yields matter when you bring pricing wars into it)


If NVidia puts this card @ $320, what happens to the 580/590? Think about that for a second. If nVidia could drop prices to sell these, the OEM`s would have by now, especially when new cards are round corner, It Won&#8217;t happen - until profits are made with newer generation cards.


&#8220;NVidia is going to be the price champion&#8221; Even so, AMD can still compete. who has better yields (AMD) R0x0rs and the less (nVidia) SuX0Rs. lol

PS Remember Die Size/Yields = Price Wars (Minus DDR &#8211; 256/384 Interface - costs etc.) I hope nVidia brings it coz I need 2 full systems soon (Gaming & HTPC).


I also hope this round isn't about, which is the premium company? lolza

 

[Elfear]

In one case a poster was implying that die sizes affect purchasing decisions by suggesting that a card with a larger die but similar performance is a turd, etc.

In my case, I am not making any claims about purchasing decisions because I don't know anything about price or AMD's possible response should they deem it necessary to lower prices. I am simply estimating performance of NV's next gen 28nm parts based on the fact that just to catch up to (beat by 20%) NV's 2-year-old architecture, it took AMD a brand new architecture + a node shrink. That makes it reasonable to estimate that the NV's brand new architecture + its own node shrink should easily surpass that mark. If not, well then NV really dropped the ball given how low the mark has been set thus far.



Agreed. That would make for an excellent clock-for-clock comparison. We don't know if GK104 is NV's mid-range or high-end though. If GK104 is only NV's upper-midrange card, that would automatically relegate HD7900 series to being upper-midrange in their performance as well, possibly paving way for a faster HD7970 to combat the faster GK110. That would be very good for consumers.



But firms aren't a charity....You have to study the technology sector more. You wait 15-18 months to upgrade your single-core smartphone to a single-core smartphone. iPad 2 turns into iPad 3 without a new HD screen and more processing power, just better power consumption. You just get new "features" and lower power consumption. You should be happy with that. Where have you been the last 10 years?

Frankly, it might not be fair to put the blame on AMD. It's not their fault that GPU consumers are OK with spending $450 for 15 months old performance level. Sounds like AMD is onto something. Maybe Read and JHH should have a meeting and time it so that each other's GPUs beat the other by 5% every 6 months and prices stay at $450-550 level forever.



I hear there is new trend in high-end GPUs: launching a next generation high-end card 30-50% underclocked from the factory. Rumor has it, GK104 will launch with 600mhz clocks at $500. Its performance will only be 5% faster than GTX580, but NV expects people to overclock it 20-30% to beat GTX580 by 25-35% to show their overclocking skills and overclocking knowledge. It makes new GPUs that much better since to extract what was previously added performance for "free" given Moore's Law, NV will also expect people to play the overclocking lottery to achieve it.

From now on, all next generation GPUs will require overclocking to achieve significantly better performance than the previous high-end GPU. The main selling features will be power consumption :sneaky:

Geez RS. I usually find your posts well thought out and interesting but AMD really seems to have gotten under your skin with the 7900 series. I (like everyone else) wish the cards had released at lower prices too but you come off like there is no value in them. I gamed on a pair of 6950's for a year and to be able to go to a single card that performed the same or better is great in my book.

Back OT - Too many rumors about the GK104. Seems the only solid ones are die size so far.

An average user that bought 5850 2 years ago, had little incentive to buy anything from AMD down the line.
And there is STILL nothing for him in AMD's basket.

Wait, what? Are you really saying that the 7970 brings no discernible performance increase over the 5850? Have you gamed on both because I have and there is no comparison.

 

[f1sherman]

@mak360

I guess you missed those $219 GTX 480 from tigerdirect and newegg

http://www.newegg.com/Product/Produc...82E16814130759

 

[f1sherman]

Wait, what? Are you really saying that the 7970 brings no discernible performance increase over the 5850?

No.

I am talking about An average user that bought 5850 2 years ago
For like $259.

I am not talking about early adopters,
who are willing to pay no-matter-what

 

[3DVagabond]

Maybe it's just an engineering sample card for testing purposes, thus the odd connectors (power and the other white ones) and the odd layout.

That's what I was thinking. There is no point to two 6's and an 8 for the final product. 2x 8pin would accomplish the same thing. Also, if the supposed midrange card required 375W worth of connectivity, what would the BigGK need?

 

[kidsafe]

@mak360

I guess you missed those $219 GTX 480 from tigerdirect and newegg

http://www.newegg.com/Product/Produc...82E16814130759

NewEgg and TigerDirect are likely selling them at barely above the price they got from their distributor (or maybe direct from EVGA). In fact it's further indication of how colossal a failure the GTX 480 and the rest of the original GF100 cards were.

Why?

- It's 2012, and high traffic retailers still have GTX 480s in stock.
- It's being sold cheaper than most GTX 560 Ti cards despite being roughly 25% faster.
- Even at that price TigerDirect did not sell out.

How much money do you think EVGA lost for having that much excess inventory?

 

[blackened23]

Looks like AIB makers are developing harvested 7970 chips that are super overclocked, sapphire was rumored to be developing a 1335 mhz 7970 with 6gb memory, and they will unveil it at cebit.

http://www.sapphiretech.com/presentation/media/media_index.aspx?psn=0004&articleID=4195&lid=1

 

[wahdangun]

Prior to AMD acquiring ATI, ATI didn't care for "small die strategy."

We got:

9700Pro (best GPU at the time)
9800Pro/9800XT (best at the time)
X800XT PE / X850XT PE (arguably better than 6800U, but can be easily considered at least as good)
X1800XT PE/X1900XT/ X1950XTX, etc. (arguably better than 7800/7900 series)
* ATI made a crapload of $, with a market cap the size of AMD.
** ATI was viewed as a premium brand.

After AMD acquired ATI and decided to focus on the small-die strategy:

2900XT (huge flop)
HD3870 (huge flop)
4870, 5870, 6970 (2nd best, relegated to competing on price).
* AMD's GPU division struggles to make $
** AMD's GPUs are viewed as a price/performance brand - brand value erosion
*** AMD's discrete GPUs continue to lose market share

There is no evidence at all that AMD's 'small die' strategy is working, not financially, not from a strong brand image perspective, not from market share. One might conclude ATI knew how to both make its videocards and manage its GPU brand MUCH better for us gamers. My gut feeling is the main reason AMD's upper-level mgmt is forcing the GPU guys to focus so much on performance/watt is because their APU strategy depends on it. NV doesn't make APUs, so they can make GPUs with as large as they need to. The end result has been NV beating AMD with 8800GTX, GTX280, GTX285, GTX480, GTX580 for 3 consecutive generations....that's got to hurt AMD's brand image in the long-term.

That's why I don't understand why people keep bringing up that 'small die size' is better. If someone wants to focus so much on advantages of smaller die sizes, they should create a separate thread.

you must be forget one tiny detail, the small die strategy let AMD regain market share, and do you forget about HD 5970 ??? that card is the fastest card for like a year and a half. and then HD 6990 is toe to toe with GTX 590.

 

[GaiaHunter]

So basically, according to you, NV hasn't made any good videocards in at least the last 5 years.

Using your logic, GTX460 was a turd since it had a larger die size than HD6850 I believe and it was slightly slower at stock speeds, but it became one of the most legendary price/performance cards in the last 5 years. HD6850 will forever be remembered as the card that existed during the time when GTX460 redefined price/performance for mid-range gamers at the time, not the other way around. No one is going to give HD3870 credit for getting smacked by 8800GT even if HD3870 did have a smaller die. Too bad. AMD should have made HD3870 with a larger die and beat 8800GT. Die size excuses don't count.

First you forget this are technical forums and so it is possible to talk about engineering which you seem to have problems with, but you have no problems with economics, as if all this isn't related.

Second you said

In one case a poster was implying that die sizes affect purchasing decisions by suggesting that a card with a larger die but similar performance is a turd, etc.

I didn't say anything about die size affecting purchase decisions. That is like if any time you talked about company profits, stocks, etc, I would said "there is a poster implying company financials affect purchase decisions".

Second, the only reason it isn't a turd in the end, is because AMD isn't going for large dies as well (I guess they need money to develop big CPU dies that are turds...).

A >500 mm^2 with 2048 mm^2 VLIW-4 shaders would be a 7970 on 40 nm, would consume as much or more power as the GTX580, but it would be a good 20% faster.

That is why, despite the GTX590 having 6000 million transistors and 1040 mm^2 of gpu silicon vs the 5280 million transistors and 778 mm^2 of silicon of the 6990, it is tit for tat.

If it was the Geforce 6850 and Geforce 6990 versus the Radeon GTX460 and Radeon GTX590, it would have been a slaughter in favour of NVIDIA, but that is due to NVIDIA being a better managed company than AMD/ATI has been.

By the way, AMD did make a slightly larger 3870, called the 4870/4850 that beat the 8800GT senseless, but came a bit too late where the 8800GT was concerned a bit like the 6870/6850 were a bit too late.

See, Management/Financials, Architecture, Timing.

In the end that turns into a GPU that we buy based on price and performance, using features (including power consumption, IQ enhancements, cooler, etc) as a sort of tie breaker.

 

[Piano Man]

http://75.126.99.220/showpost.php?p=1038447903&postcount=437

Ok, now I'm getting excited. If I can get 7970 performance (or better) for 300, I'm not even going to hesitate to go over to team green this round.

 

[PingviN]

After AMD acquired ATI and decided to focus on the small-die strategy:

2900XT (huge flop)
HD3870 (huge flop)
4870, 5870, 6970 (2nd best, relegated to competing on price).
* AMD's GPU division struggles to make $
** AMD's GPUs are viewed as a price/performance brand - brand value erosion
*** AMD's discrete GPUs continue to lose market share

AMD acquired ATi in 2006, you think AMD really had much to do with HD2900 which was released in 2007? I'm guessing HD2900 was pretty much set in stone and getting ready for manufacturing shortly after the acquisition. It also wasn't a small GPU, it was a pretty big one. Just wasn't bigger than G80.
HD3870 was a follow-up. Again, probably not much AMD could do that ATi wasn't already planning. I'm guessing, of course.

 

[Crap Daddy]

Regarding die-sizes, you know that GK110 will be in excess of 500mm2? This is the high-end chip which was supposed to compete with Tahiti. If this small die GK104 will compete with AMD's flagship you do realize that Charlie's statement "Kepler vs. GCN a clear winner" is correct.

 

[GaiaHunter]

Regarding die-sizes, you know that GK110 will be in excess of 500mm2? This is the high-end chip which was supposed to compete with Tahiti. If this small die GK104 will compete with AMD's flagship you do realize that Charlie's statement "Kepler vs. GCN a clear winner" is correct.

This small die GK104 is rumoured to be similar size to the AMD flagship.

So if we are comparing absolute performances that might be the case, if/when NVIDIA brings that 500 mm^2 gpu with no response from AMD.

Kepler vs GCN seems to be an architectural argument and that is harder to say if chips have similar size/transistor count.

Charlie also said that he thought GK104 was quite smaller than Tahiti, which the more recent rumours don't seem to indicate.

 

[railven]

Just to make the point clear, if GK104 is NV's highest end but it only matches HD7970 in performance and costs $550-600, I would still consider both not living up to expectations. By extension, that would mean the entire 28nm generation will only give us what normally would be considered upper-midrange levels of performance but at high-end GPU pricing. I don't believe this will happen. I think NV's highest end single GPU will be much faster than a stock HD7970.

I get you guy, and where I disagree (at this point) is the performance increase. The only issue with the performance increase people seem to be having seems to be the cost.

The only difference this generation saw versus prior generations was that the GTX 580 saw absolutely zero price cuts. Not a singl MSRP price reduction since it's launch. So the price ceiling was already set high.

This ties into your next post:

But firms aren't a charity....You have to study the technology sector more. You wait 15-18 months to upgrade your single-core smartphone to a single-core smartphone. iPad 2 turns into iPad 3 without a new HD screen and more processing power, just better power consumption. You just get new "features" and lower power consumption. You should be happy with that. Where have you been the last 10 years?

Frankly, it might not be fair to put the blame on AMD. It's not their fault that GPU consumers are OK with spending $450 for 15 months old performance level. Sounds like AMD is onto something. Maybe Read and JHH should have a meeting and time it so that each other's GPUs beat the other by 5% every 6 months and prices stay at $450-550 level forever.

I agree with your first paragraph, but find the second paragraph ironic. People were buying 12"-month old performance levels" when they were buying GTX 580s. The card saw no price drop, clearly nVidia showed people would pay for these products. AMD followed suit. In the end of course the consumers are to blame, but not sure how many consumers were buying $500 AMD cards between Nov-2010 and Feb-2012.

Just ironic how we point fingers around here.

 

[BallaTheFeared]

This small die GK104 is rumoured to be similar size to the AMD flagship.

So if we are comparing absolute performances that might be the case, if/when NVIDIA brings that 500 mm^2 gpu with no response from AMD.

Kepler vs GCN seems to be an architectural argument and that is harder to say if chips have similar size/transistor count.

Charlie also said that he thought GK104 was quite smaller than Tahiti, which the more recent rumours don't seem to indicate.

I thought these leaks showed a die that was smaller than Tahiti.

 

[Silverforce11]

Its about the same size.

Looks like March 23rd launch.. which means we will see some performance leaks very soon.