How/why did Amd/ATI fall behind Nvidia?

[Cookie Monster]

It could had worked. But the major sticking point was Jen-Hsun Huang wanted to run both companies. Ruiz couldnt have that. He still had several years of imploding AMD left in him.

If Ruiz steps aside and Jen-Hsun Huang runs the joint company. It would have a lot better chance than they had.

I only say this because both of them are green.. joking aside nVIDIA had fantastic chipsets for AMD (not quite so for intel..). It felt more natural for them to join and id favor JHH leading the company as opposed to Ruiz. Would have been a completely different AMD.

Just think Opertons and Tesla/Quadro card platform. The server/workstation would have seen a huge change from what we see now and how they'd push HPC/GPGPU.

 

[Cookie Monster]

For all the people saying that AMD is still competitive: look at the GPU sizes. Traditionally, AMD was able to compete with Nvidia using similar die sizes, sometimes even smaller. That's very important from a business perspective, because larger dice cost more money, use more power, and usually also have wider memory buses (which increases the cost of the PCB and auxiliary components).

Against Kepler, AMD was pretty competitive. Tahiti (352 sq.mm., 4.3 billion transistors) was a bit bigger and more power-hungry than GK104 (294 sq.mm., 3.5 billion transistors), but it wasn't a huge difference and was partially justified by Tahiti's better compute performance (especially in Double Precision). These chips were at least in the same basic class. On the extra-large dice, AMD came out ahead; 438 sq.mm. Hawaii (6.2 billion transistors) was a better gaming chip than the 561 sq.mm. GK110 (7.1 billion transistors) and was competitive in GPGPU (including Double Precision performance) as well. That said, even though AMD had the technical advantage on this battle, Nvidia won the business competition by a mile; they made far more on GK110 Tesla than AMD could dream of doing on FirePro Hawaii, and on the consumer side, they first milked the high-end market for $999 Titan sales, and then beat AMD on the midrange in large part simply by providing a stock cooler that didn't totally suck. AMD really botched the Hawaii release badly, and it hurt them a lot. IMO, it should have been a pure "virtual" release (no reference cards), and the R9 290 should have had a lower power limit for more energy efficiency since unlike the R9 290X it didn't need to win the performance crown outright. Another example of AMD's competent engineers being let down by their abysmal marketing team.

Against Maxwell, though... AMD simply can't compete at all if you look at what kind of chips they have to price against Nvidia's offerings to retain even what little market share they have. Consider the following: Pitcairn is a 212mm^2 chip with 2.8 billion transistors. GM206 is a 227mm^2 chip with 2.9 billion transistors. These chips should be in roughly the same performance and price bracket - but they're not even close, since AMD is still using an obsolete 2012 design. AMD is lucky to get $130-$150 for a Pitcairn card, while Nvidia's GM206-based GTX 960 is selling like hotcakes at $200-$240 (depending on RAM, featureset, etc.) And it's even worse than that, because the GM206 only has a 128-bit bus while Pitcairn has a 256-bit bus. So the AMD boards are more expensive to produce: more traces and more RAM chips are required. To compete with GM206 in terms of performance, AMD has to use Tonga - a much larger chip (359 mm^2, 5.0 billion transistors) with higher power requirements. They therefore sell fewer cards at the same price point, and make less profit on those they do sell. And that means fewer R&D dollars for the next generation (if it ever comes).

Likewise, we've heard people say that Hawaii can compete with GM204 (GTX 970/GTX 980) in terms of performance. Sure, if the only thing you're concerned about is raw framerates in today's AAA games, it can. But in every other metric (except Double Precision, where it still competes with Kepler), Hawaii falls way behind. GM204 is a 398 sq.mm. chip with 5.2 billion transistors: not much bigger than Tonga. Yet the Tonga-based R9 285 has less than 65% of the GM204-based GTX 970's performance. To get that performance level, AMD has to resort to Hawaii, with a 438 sq.mm. chip that contains 6.2 billion transistors - and they have to push it far beyond its optimal performance per watt, which hurts sales. Hawaii has a massive 512-bit memory bus, and it needs a very robust power stage, which makes boards expensive to design. Profits on Hawaii must be tiny compared to profits on GM204. Again, this means less money to sustain AMD's development for newer graphics chips.

Well said.

 

[DDH]

For all the people saying that AMD is still competitive: look at the GPU sizes. Traditionally, AMD was able to compete with Nvidia using similar die sizes, sometimes even smaller. That's very important from a business perspective, because larger dice cost more money, use more power, and usually also have wider memory buses (which increases the cost of the PCB and auxiliary components).

Against Kepler, AMD was pretty competitive. Tahiti (352 sq.mm., 4.3 billion transistors) was a bit bigger and more power-hungry than GK104 (294 sq.mm., 3.5 billion transistors), but it wasn't a huge difference and was partially justified by Tahiti's better compute performance (especially in Double Precision). These chips were at least in the same basic class. On the extra-large dice, AMD came out ahead; 438 sq.mm. Hawaii (6.2 billion transistors) was a better gaming chip than the 561 sq.mm. GK110 (7.1 billion transistors) and was competitive in GPGPU (including Double Precision performance) as well. That said, even though AMD had the technical advantage on this battle, Nvidia won the business competition by a mile; they made far more on GK110 Tesla than AMD could dream of doing on FirePro Hawaii, and on the consumer side, they first milked the high-end market for $999 Titan sales, and then beat AMD on the midrange in large part simply by providing a stock cooler that didn't totally suck. AMD really botched the Hawaii release badly, and it hurt them a lot. IMO, it should have been a pure "virtual" release (no reference cards), and the R9 290 should have had a lower power limit for more energy efficiency since unlike the R9 290X it didn't need to win the performance crown outright. Another example of AMD's competent engineers being let down by their abysmal marketing team.

Against Maxwell, though... AMD simply can't compete at all if you look at what kind of chips they have to price against Nvidia's offerings to retain even what little market share they have. Consider the following: Pitcairn is a 212mm^2 chip with 2.8 billion transistors. GM206 is a 227mm^2 chip with 2.9 billion transistors. These chips should be in roughly the same performance and price bracket - but they're not even close, since AMD is still using an obsolete 2012 design. AMD is lucky to get $130-$150 for a Pitcairn card, while Nvidia's GM206-based GTX 960 is selling like hotcakes at $200-$240 (depending on RAM, featureset, etc.) And it's even worse than that, because the GM206 only has a 128-bit bus while Pitcairn has a 256-bit bus. So the AMD boards are more expensive to produce: more traces and more RAM chips are required. To compete with GM206 in terms of performance, AMD has to use Tonga - a much larger chip (359 mm^2, 5.0 billion transistors) with higher power requirements. They therefore sell fewer cards at the same price point, and make less profit on those they do sell. And that means fewer R&D dollars for the next generation (if it ever comes).

Likewise, we've heard people say that Hawaii can compete with GM204 (GTX 970/GTX 980) in terms of performance. Sure, if the only thing you're concerned about is raw framerates in today's AAA games, it can. But in every other metric (except Double Precision, where it still competes with Kepler), Hawaii falls way behind. GM204 is a 398 sq.mm. chip with 5.2 billion transistors: not much bigger than Tonga. Yet the Tonga-based R9 285 has less than 65% of the GM204-based GTX 970's performance. To get that performance level, AMD has to resort to Hawaii, with a 438 sq.mm. chip that contains 6.2 billion transistors - and they have to push it far beyond its optimal performance per watt, which hurts sales. Hawaii has a massive 512-bit memory bus, and it needs a very robust power stage, which makes boards expensive to design. Profits on Hawaii must be tiny compared to profits on GM204. Again, this means less money to sustain AMD's development for newer graphics chips.

hawaii is 2013 tech, wasnt released with anticipation of competing against a 2014 refreshed nvidia line. AMD didnt answer GM204. Whether it is because the 20nm plans fell through or for some other unknown reason. So an analysis of hawaii v gm204 in terms of size is meaningless unless you acknowledge that hawaii is a generation before gm204

Fury v GM200 is a proper comparison. GM200 is 601mm sq. If AMD come in under, with on par performance then your conclusions are for naught.

 

[gamervivek]

And apparently Maxwell appeared out of ether? If GCN is still competitive, AMD only have to get something to close the clockspeed deficit, then why should they care?

 

[Red Hawk]

For all the people saying that AMD is still competitive: look at the GPU sizes. Traditionally, AMD was able to compete with Nvidia using similar die sizes, sometimes even smaller. That's very important from a business perspective, because larger dice cost more money, use more power, and usually also have wider memory buses (which increases the cost of the PCB and auxiliary components).

Against Kepler, AMD was pretty competitive. Tahiti (352 sq.mm., 4.3 billion transistors) was a bit bigger and more power-hungry than GK104 (294 sq.mm., 3.5 billion transistors), but it wasn't a huge difference and was partially justified by Tahiti's better compute performance (especially in Double Precision). These chips were at least in the same basic class. On the extra-large dice, AMD came out ahead; 438 sq.mm. Hawaii (6.2 billion transistors) was a better gaming chip than the 561 sq.mm. GK110 (7.1 billion transistors) and was competitive in GPGPU (including Double Precision performance) as well. That said, even though AMD had the technical advantage on this battle, Nvidia won the business competition by a mile; they made far more on GK110 Tesla than AMD could dream of doing on FirePro Hawaii, and on the consumer side, they first milked the high-end market for $999 Titan sales, and then beat AMD on the midrange in large part simply by providing a stock cooler that didn't totally suck. AMD really botched the Hawaii release badly, and it hurt them a lot. IMO, it should have been a pure "virtual" release (no reference cards), and the R9 290 should have had a lower power limit for more energy efficiency since unlike the R9 290X it didn't need to win the performance crown outright. Another example of AMD's competent engineers being let down by their abysmal marketing team.

Against Maxwell, though... AMD simply can't compete at all if you look at what kind of chips they have to price against Nvidia's offerings to retain even what little market share they have. Consider the following: Pitcairn is a 212mm^2 chip with 2.8 billion transistors. GM206 is a 227mm^2 chip with 2.9 billion transistors. These chips should be in roughly the same performance and price bracket - but they're not even close, since AMD is still using an obsolete 2012 design. AMD is lucky to get $130-$150 for a Pitcairn card, while Nvidia's GM206-based GTX 960 is selling like hotcakes at $200-$240 (depending on RAM, featureset, etc.) And it's even worse than that, because the GM206 only has a 128-bit bus while Pitcairn has a 256-bit bus. So the AMD boards are more expensive to produce: more traces and more RAM chips are required. To compete with GM206 in terms of performance, AMD has to use Tonga - a much larger chip (359 mm^2, 5.0 billion transistors) with higher power requirements. They therefore sell fewer cards at the same price point, and make less profit on those they do sell. And that means fewer R&D dollars for the next generation (if it ever comes).

Likewise, we've heard people say that Hawaii can compete with GM204 (GTX 970/GTX 980) in terms of performance. Sure, if the only thing you're concerned about is raw framerates in today's AAA games, it can. But in every other metric (except Double Precision, where it still competes with Kepler), Hawaii falls way behind. GM204 is a 398 sq.mm. chip with 5.2 billion transistors: not much bigger than Tonga. Yet the Tonga-based R9 285 has less than 65% of the GM204-based GTX 970's performance. To get that performance level, AMD has to resort to Hawaii, with a 438 sq.mm. chip that contains 6.2 billion transistors - and they have to push it far beyond its optimal performance per watt, which hurts sales. Hawaii has a massive 512-bit memory bus, and it needs a very robust power stage, which makes boards expensive to design. Profits on Hawaii must be tiny compared to profits on GM204. Again, this means less money to sustain AMD's development for newer graphics chips.

This performance difference probably comes down to the color compression algorithms Nvidia has implemented with Maxwell. Tonga has color compression, but not as good as Maxwell's. AMD really needs to refresh their whole product line and come up with an answer to that color compression. HBM just might be that answer, but rumors have it that AMD is only introducing new products at the very high end, replacing Tahiti with Tonga, and keeping around Hawaii, Pitcairn, Bonaire, and Oland. That's really not competitive with how Nvidia has Maxwell products from top to bottom. In the past AMD has been pretty good about getting new chips out after a few years on the same process -- see the four different chips they introduced for the 6000 series, Cayman, Barts, Turks, and Caicos. They're riding their GCN cards out as long as the can. The only real reason that I can imagine they're doing this is that they're at least running low on R&D resources, and they simply can't afford to refresh their entire product stack with new chips, even if the current chips just can't compete. It's sad, really.

 

[xthetenth]

ANd what the old saying..........?

"You get what you pay for".

Sometimes what you pay for is the satisfaction of having bought the more expensive product, no more.

 

[Cookie Monster]

And apparently Maxwell appeared out of ether? If GCN is still competitive, AMD only have to get something to close the clockspeed deficit, then why should they care?

In terms of performance perhaps. In terms of everything else no e.g. die size, no of transistors, power efficiency which ultimately determines cooling/VRM BOM cost/PCB complexity, 512bit vs 256bit (need more amount of memory IC on top of a more robust power circuitry, real estate etc) etc.

 

[sandorski]

In terms of performance perhaps. In terms of everything else no e.g. die size, no of transistors, power efficiency which ultimately determines cooling/VRM BOM cost/PCB complexity, 512bit vs 256bit (need more amount of memory IC on top of a more robust power circuitry, real estate etc) etc.

All those factors, except Power Consumption, have no affect on the Consumer's experience.

 

[JDG1980]

All those factors, except Power Consumption, have no affect on the Consumer's experience.

Sure they can. If you have a mini-ITX case (and small form factors are increasingly popular), then a big PCB as required by Hawaii can be a compatibility problem. In fairness, AMD seems to recognize that this is an issue, and the smaller Fury PCB is an attempt to address it that may well be successful.

And power consumption is a big deal; one reason why Nvidia's GTX 750 Ti is such a big seller (and was a smart first choice for the Maxwell launch) is that this card can be stuck into almost any PC with a PCIe x16 slot, even cheap OEM boxes with crappy power supplies.

 

[gamervivek]

In terms of performance perhaps. In terms of everything else no e.g. die size, no of transistors, power efficiency which ultimately determines cooling/VRM BOM cost/PCB complexity, 512bit vs 256bit (need more amount of memory IC on top of a more robust power circuitry, real estate etc) etc.

The 398 figure for GM204 is a nvidia figure, TR measured it at 416mm2, so not that big of a difference. Power efficiency and higher clocks are joined at hip. Color compression will take care of the bus-width.

AMD didn't magically outcompete nvidia with smaller dies, they had them running at higher clocks. And they'd still be behind nvidia's top chip by more than 290X is behind 980 at 4k and it's running Kepler ragged.

 

[sandorski]

Sure they can. If you have a mini-ITX case (and small form factors are increasingly popular), then a big PCB as required by Hawaii can be a compatibility problem. In fairness, AMD seems to recognize that this is an issue, and the smaller Fury PCB is an attempt to address it that may well be successful.

And power consumption is a big deal; one reason why Nvidia's GTX 750 Ti is such a big seller (and was a smart first choice for the Maxwell launch) is that this card can be stuck into almost any PC with a PCIe x16 slot, even cheap OEM boxes with crappy power supplies.

You are really reaching there. The size of the GPU itself has no affect on the size of the PCB.

 

[Cookie Monster]

All those factors, except Power Consumption, have no affect on the Consumer's experience.

We aren't talking consumer experience are we? I agree that for us it doesn't matter as long as we get decent frames without a jet taking off. However financially.. AMD are bleeding money and the marketshare is tipping to 70/30 (or was it 75/25) which in the long run will not be good for the consumer because they'll spend less and less on R&D to compensate or just concede defeat.. Its not a charity. And you want to perform the best as cheaply as possible to turn a profit. This applies to almost every product out there.

 

[JDG1980]

You are really reaching there. The size of the GPU itself has no affect on the size of the PCB.

Yes, in theory, you can put a large GPU on a tiny PCB (and in fact, this is what AMD is doing with the watercooled Fury). But you can't do this with Hawaii, because Hawaii needs 16 GDDR5 RAM chips (and the PCB traces to route them), plus a massive power stage.

Just take a look at all this crap. How much can this board really be shrunk down?

You'll never see a card like this or this with Hawaii. Note that AMD can do it with their latest GCN 1.2 design, but not with the old rebrands they're trotting out.

 

[Genx87]

Sure they can. If you have a mini-ITX case (and small form factors are increasingly popular), then a big PCB as required by Hawaii can be a compatibility problem. In fairness, AMD seems to recognize that this is an issue, and the smaller Fury PCB is an attempt to address it that may well be successful.

And power consumption is a big deal; one reason why Nvidia's GTX 750 Ti is such a big seller (and was a smart first choice for the Maxwell launch) is that this card can be stuck into almost any PC with a PCIe x16 slot, even cheap OEM boxes with crappy power supplies.

Heh I had a 5870 that required I use a hammer to bend a drive cage to get into an Antec Sonata II Mid ATX case lol.

 

[chimaxi83]

Hey, PCB size, another thing to add to the "made up crap to whine about AMD" list. What's next, AIO tubing? Oh wait, that's already there. Illuminated logo, then? No, that's covered. How about name? Nope, whiners have already let it be known that Fury is too violent. Maybe color scheme is next?

 

[sandorski]

We aren't talking consumer experience are we? I agree that for us it doesn't matter as long as we get decent frames without a jet taking off. However financially.. AMD are bleeding money and the marketshare is tipping to 70/30 (or was it 75/25) which in the long run will not be good for the consumer because they'll spend less and less on R&D to compensate or just concede defeat.. Its not a charity. And you want to perform the best as cheaply as possible to turn a profit. This applies to almost every product out there.

When Marketshare is the argument, then yes we are talking about the Consumer. They determine Marketshare.

 

[sandorski]

Yes, in theory, you can put a large GPU on a tiny PCB (and in fact, this is what AMD is doing with the watercooled Fury). But you can't do this with Hawaii, because Hawaii needs 16 GDDR5 RAM chips (and the PCB traces to route them), plus a massive power stage.

Just take a look at all this crap. How much can this board really be shrunk down?

You'll never see a card like this or this with Hawaii. Note that AMD can do it with their latest GCN 1.2 design, but not with the old rebrands they're trotting out.

....and? Nvidia has all the same on their cards. It has absolutely nothing to do with the size of the GPU.

 

[Wreckem]

Sure they can. If you have a mini-ITX case (and small form factors are increasingly popular), then a big PCB as required by Hawaii can be a compatibility problem. In fairness, AMD seems to recognize that this is an issue, and the smaller Fury PCB is an attempt to address it that may well be successful.

And power consumption is a big deal; one reason why Nvidia's GTX 750 Ti is such a big seller (and was a smart first choice for the Maxwell launch) is that this card can be stuck into almost any PC with a PCIe x16 slot, even cheap OEM boxes with crappy power supplies.

Um. The size of 290x has rarely been a problem for mitx machines. There are people who have crammed 295x2 into true mitx machines.

 

[3DVagabond]

ANd what the old saying..........?

"You get what you pay for".

Like Titan vs. 780ti, Titan Z vs. a whole PC included with a 295x2, Titan X vs 980 ti? Face it, sometimes you just get screwed over.

 

[3DVagabond]

For all the people saying that AMD is still competitive: look at the GPU sizes. Traditionally, AMD was able to compete with Nvidia using similar die sizes, sometimes even smaller. That's very important from a business perspective, because larger dice cost more money, use more power, and usually also have wider memory buses (which increases the cost of the PCB and auxiliary components).

Against Kepler, AMD was pretty competitive. Tahiti (352 sq.mm., 4.3 billion transistors) was a bit bigger and more power-hungry than GK104 (294 sq.mm., 3.5 billion transistors), but it wasn't a huge difference and was partially justified by Tahiti's better compute performance (especially in Double Precision). These chips were at least in the same basic class. On the extra-large dice, AMD came out ahead; 438 sq.mm. Hawaii (6.2 billion transistors) was a better gaming chip than the 561 sq.mm. GK110 (7.1 billion transistors) and was competitive in GPGPU (including Double Precision performance) as well. That said, even though AMD had the technical advantage on this battle, Nvidia won the business competition by a mile; they made far more on GK110 Tesla than AMD could dream of doing on FirePro Hawaii, and on the consumer side, they first milked the high-end market for $999 Titan sales, and then beat AMD on the midrange in large part simply by providing a stock cooler that didn't totally suck. AMD really botched the Hawaii release badly, and it hurt them a lot. IMO, it should have been a pure "virtual" release (no reference cards), and the R9 290 should have had a lower power limit for more energy efficiency since unlike the R9 290X it didn't need to win the performance crown outright. Another example of AMD's competent engineers being let down by their abysmal marketing team.

Against Maxwell, though... AMD simply can't compete at all if you look at what kind of chips they have to price against Nvidia's offerings to retain even what little market share they have. Consider the following: Pitcairn is a 212mm^2 chip with 2.8 billion transistors. GM206 is a 227mm^2 chip with 2.9 billion transistors. These chips should be in roughly the same performance and price bracket - but they're not even close, since AMD is still using an obsolete 2012 design. AMD is lucky to get $130-$150 for a Pitcairn card, while Nvidia's GM206-based GTX 960 is selling like hotcakes at $200-$240 (depending on RAM, featureset, etc.) And it's even worse than that, because the GM206 only has a 128-bit bus while Pitcairn has a 256-bit bus. So the AMD boards are more expensive to produce: more traces and more RAM chips are required. To compete with GM206 in terms of performance, AMD has to use Tonga - a much larger chip (359 mm^2, 5.0 billion transistors) with higher power requirements. They therefore sell fewer cards at the same price point, and make less profit on those they do sell. And that means fewer R&D dollars for the next generation (if it ever comes).

Likewise, we've heard people say that Hawaii can compete with GM204 (GTX 970/GTX 980) in terms of performance. Sure, if the only thing you're concerned about is raw framerates in today's AAA games, it can. But in every other metric (except Double Precision, where it still competes with Kepler), Hawaii falls way behind. GM204 is a 398 sq.mm. chip with 5.2 billion transistors: not much bigger than Tonga. Yet the Tonga-based R9 285 has less than 65% of the GM204-based GTX 970's performance. To get that performance level, AMD has to resort to Hawaii, with a 438 sq.mm. chip that contains 6.2 billion transistors - and they have to push it far beyond its optimal performance per watt, which hurts sales. Hawaii has a massive 512-bit memory bus, and it needs a very robust power stage, which makes boards expensive to design. Profits on Hawaii must be tiny compared to profits on GM204. Again, this means less money to sustain AMD's development for newer graphics chips.

How can you even bring up DP while trying to talk nVidia being better than AMD. Current nVidia acrh is absolutely horrendous with DP. I doubt it could compete with Fermi possibly not even VLIW chips never mind GCN.

 

[SolMiester]

Those who paid more money for a 680 and 770 over 7970/280X or 780 over 290 sure didn't.

My biggest disappointment with humanity in so far as GPUs goes is when people disappointed in Kepler performance just upgrade to Maxwell instead of trying to change. Actually, I take that back. Those that taught Nvidia a lesson by returning their 970s and buying a 980 when they found out about its true config. That learned 'em.

Why dont you take us throught this then....7970 came out at high than normal price for AMD, by with broken CF drivers for first 6mths, then followed 2 yrs until FCATs was acceptable on CF 7xxx, but agin, excluding >1600p and not DX9 CF.
680 came out at same price as previous x80 series card, about $50 cheaper, and at time, faster.
Meanwhile, NV then re-releases keplar to 770 and 79xx catches up with 680 performance but still no fixed CF drivers.
780 came out and took crown again, 290 released, takes crown 780Ti out, takes crown back......
290 on release got sucked up by miners, so people paid inflated prices anyway. AMD has nothing more to release so has spent last 12 mths improving drivers....well, if thats all you have to play with, I would hope they would do their best to improve performance.
Fact of the matter is, neither 680, 290 ot 780's are good enough for 4k, so no GPU is out in front of the others.
Im struggling to see your point.

 

[SolMiester]

How can you even bring up DP while trying to talk nVidia being better than AMD. Current nVidia acrh is absolutely horrendous with DP. I doubt it could compete with Fermi possibly not even VLIW chips never mind GCN.

He was excluding DP..!...

 

[caswow]

He was excluding DP..!...

you cant exclude dp and at the same time say amd has an inferior gpu design.

 

[Genx87]

you cant exclude dp and at the same time say amd has an inferior gpu design.

We can if all we are doing is playing video games.

 

[Headfoot]

Title assumes a conclusion without providing any logic to back it up? Check.
Easily identifiable, hard factual qualifications to prevent flame war easily available yet not utilized? (E.g. Why is AMD trailing in market share)? up? Check
A few dozen word long OP? up? Check

Diagnosis: Flame Bait Thread.