[BitsAndChips]390X ready for launch - AMD ironing out drivers - Computex launch

[3DVagabond]

You are missing the point.

AMD makes smaller GPUs that come close to bigger NV GPU. The performance gap has been historically 15-20%.

The gap closed a great deal with Hawaii vs GK110, with the R290X beating Titan and eventually beating 780Ti.

Analyze it on a tech level. Die size, TDP, architecture, ram advantage. If AMD's latest 550mm2 die cannot beat GM200, it's a failure.

I understand what you are saying. It will be a technical failure. With what AMD has done in the past and considering the size and HBM, AMD should beat nVidia this round performance wise.

Now to what Russian Sensation is saying, If fast enough and priced right it still could be a commercial success though.

 

[Silverforce11]

I understand what you are saying. It will be a technical failure. With what AMD has done in the past and considering the size and HBM, AMD should beat nVidia this round performance wise.

Now to what Russian Sensation is saying, If fast enough and priced right it still could be a commercial success though.

I'm trying to point out that AMD hasn't ever been a commercial success when its slower and priced lower. It's a slow road to death.

 

[RussianSensation]

You are missing the point.

AMD makes smaller GPUs that come close to bigger NV GPU. The performance gap has been historically 15-20%.

The gap closed a great deal with Hawaii vs GK110, with the R290X beating Titan and eventually beating 780Ti.

Analyze it on a tech level. Die size, TDP, architecture, ram advantage. If AMD's latest 550mm2 die cannot beat GM200, it's a failure.

Your point only makes sense if I am an engineer who designed R9 390X and am competing for a top engineering position at NV, or vice versa. Since I am not an engineer, but a consumer, an R9 390x could be 1mm2 or 5000mm2; what do I care as far as gaming performance goes and how much it costs me to buy the product?

As long as the card is cool and quiet and has awesome price/performance, what do I care how big its die size is? Generally speaking, we look at die size as a fun metric to try to extrapolate performance of some new unreleased card. However, once the card is actually inside the PC, I couldn't care less about what die size it is, what node it's made on, if it has GDDR5 or HBM, etc. If you can manage to make a card on a 180nm node with DDR1 that beats Titan X at a reasonable price, I'll buy it.

I'm trying to point out that AMD hasn't ever been a commercial success when its slower and priced lower. It's a slow road to death.

If you are going to argue that point, why focus so much on desktop graphics when (1) market growth (2) profit margins (3) greater potential for market share gains -- all lie in the notebook GPU market. I said from the beginning if we just want to discuss AMD as a business, they should not even make R9 390 series but put most of their efforts towards mobile dGPUs. You didn't like this idea. If we are going to prioritize the importance for AMD, it's a no brainer to vote notebook dGPU over desktop. At that point AMD would have been way better off focusing its efforts on designing the world's most efficient mobile dGPUs and sacrifice desktop GPUs until 14nm.

 

[Silverforce11]

Because if AMD can't beat NV with a similar huge die size (when Hawaii vs GK110 shows they can match or exceed with a 25% smaller chip) with their best architecture AND a massive vram advantage, they have no chance when NV gets onto Pascal + HBM. Period.

AMD isn't going to be selling many 390X at $599 if its slower than Titan X. Why? Cos the 980Ti is coming, as fast or faster than Titan X at ~$799. The gap will be a blow out. They'll be forced to sell 390X for $399, then the poor 390 also with HBM, at what, $299? lol. Can't imagine there's much profits to be had with expensive HBM and such a large GPU.

Importantly, a factor you need to acknowledge whether you like it or not: The market have indeed spoken (thx SirPauly), gamers aren't willing to pay top $ for AMD GPU unless its got the same performance & similar efficiency as NV, re: R290/X at $400/$500 vs 780/Ti at $500/600 (that $100 difference is the NV tax gamers are willing to pay). If AMD wants to sell 390X at $799, it needs to be faster than Titan X because its competitor isn't Titan X, it's the 980Ti at $799.

I'm focused on desktop dGPU because AMD's notebook dGPU is AWOL for years. It's a lost market until they can cause a revolution with large APU + HBM combo.

 

[raghu78]

Historically, AMD has never been so close on performance to NV's big-die as Hawaii vs GK110 was/is (compare 5870 vs 480, 6970 vs 580, Tahiti vs GK110). It's a steady progression where AMD is catching up. For them to suddenly go balls to wall with a massive GPU + new vram tech and not exceed NV's big-die is pure fail.

NV's Maxwell chips throw fp64 under the bus and thats why GM200 is 600 sq mm. Nvidia's increased performance efficiency (perf /sqmm) came at the cost of gutting fp64 performance. GM200's fp64 performance is a pitiable 200 - 220 GFLOPS.

http://www.anandtech.com/show/9096/nvidia-announces-quadro-m6000-quadro-vca-2015

On the contrary next gen AMD flagship Firepros will keep the tradition with the same 1/2 fp64 rate as Hawaii based Firepro W9000. the flagship Firepros based on the next gen GCN flagship will have a stunning 3.5 TFLOPS of fp64 performance even accounting for lower clocks of 900 Mhz for staying within 250W TDP.

The next flagship Tesla product from Nvidia will not launch till Q2 2017 when GP100 launches. Thats the reason why Nvidia launched dual GK210 based Tesla K80 in Nov 2014.

http://www.anandtech.com/show/8729/nvidia-launches-tesla-k80-gk210-gpu

Even assuming AMD launches their next gen flagship Firepro in Q4 2015 or early Q1 2016 they will have no competition to their fp64 single GPU monster for the full year 2016 in HPC.

I do not expect Nvidia to have a smooth transition to 16/14nm, Pascal, NVlink and HBM. If they go direct to an all out transition without any Maxwell 16/14nm shrinks they are taking on even more risk than Fermi GF100. Lets see how this plays out. Nvidia's strength is GPU architecture and software. AMD's strength is engineering and thats the reason AMD was first to GDDR5 and have now co-invented HBM along with Hynix and proposed it to JEDEC which adopted it as a standard JESD235 in Oct 2013. AMD's efforts with HBM will pay over the next 2 years just as it happened with the HD 4870 to HD 5000 series.

 

[Silverforce11]

AMD's efforts with HBM will pay over the next 2 years just as it happened with the HD 4870 to HD 5000 series.

Good post for the HPC segment. Indeed the 390X may make for an excellent professional GPU.

I agree with most of your points except the last.

AMD had the lead with GDDR5, didn't give them any advantage since NV caught up and eventually made a better MC that can push higher vram clocks.

 

[RussianSensation]

AMD isn't going to be selling many 390X at $599 if its slower than Titan X. Why? Cos the 980Ti is coming, as fast or faster than Titan X at ~$799. The gap will be a blow out. They'll be forced to sell 390X for $399, then the poor 390 also with HBM, at what, $299? lol. Can't imagine there's much profits to be had with expensive HBM and such a large GPU.

Oh, we are back to the argument that the only market that exists are $799+ GPUs. I forgot. Also, we have no valid info on 980Ti, no performance data, no price data, but you already know its price and performance.

If AMD wants to sell 390X at $799, it needs to be faster than Titan X because its competitor isn't Titan X, it's the 980Ti at $799.

Why would you think AMD is targeting $799 cuz some random person on the Internet started this unsubstantiated rumour? Unless someone here works for AMD, we have no clue what their intended launch price is. Just like 1 person online said 980Ti will cost $799 as a fact and everyone believed it? AT forums = where opinions and conjecture are automatically facts! :whiste:

NV's Maxwell chips throw fp64 under the bus and thats why GM200 is 600 sq mm. Nvidia's increased performance efficiency (perf /sqmm) came at the cost of gutting fp64 performance. GM200's fp64 performance is a pitiable 200 - 220 GFLOPS.

http://www.anandtech.com/show/9096/nvidia-announces-quadro-m6000-quadro-vca-2015

If 390X has 10-20x the DP performance of the Titan X, it'll be ignored. Instead the focus will fall on the extra 50-60W of power it uses. You shouldn't be suprised though because all the original Titan owners who used DP to justify the price premium no longer care about DP since Maxwell bombs in it. Don't you worry, once Pascal has DP in spades, esp. if it beats AMD at this metric, the pro-DP discussion will be back on the table!

 

[raghu78]

AMD isn't going to be selling many 390X at $599 if its slower than Titan X. Why? Cos the 980Ti is coming, as fast or faster than Titan X at ~$799. The gap will be a blow out. They'll be forced to sell 390X for $399, then the poor 390 also with HBM, at what, $299? lol. Can't imagine there's much profits to be had with expensive HBM and such a large GPU.

you are wildly guessing. You think Nvidia will have no salvage SKU on a massive 600 sq mm GM200 chip which is at the limit of 28nm TSMC node when they have a salvage GTX 970 for a 398 sq mm GM200 chip. lets just i disagree. Until we see GTX 980 Ti specs and performance on launch day maybe you should reserve your AMD will be destroyed theme. :whiste:

Importantly, a factor you need to acknowledge whether you like it or not: The market have indeed spoken (thx SirPauly), gamers aren't willing to pay top $ for AMD GPU unless its got the same performance & similar efficiency as NV, re: R290/X at $400/$500 vs 780/Ti at $500/600 (that $100 difference is the NV tax gamers are willing to pay). If AMD wants to sell 390X at $799, it needs to be faster than Titan X because its competitor isn't Titan X, it's the 980Ti at $799.

I'm focused on desktop dGPU because AMD's notebook dGPU is AWOL for years. It's a lost market until they can cause a revolution with large APU + HBM combo.

For the nth time do not argue with confidence on unannounced products without a clue as to specs and performance.

 

[raghu78]

Good post for the HPC segment. Indeed the 390X may make for an excellent professional GPU.

I agree with most of your points except the last.

AMD had the lead with GDDR5, didn't give them any advantage since NV caught up and eventually made a better MC that can push higher vram clocks.

Sorry but you need to refresh your memory. AMD started making gains in both desktop and notebook GPUs starting with HD 4800 series and consolidated it with HD 5000 series

http://www.xbitlabs.com/news/graphi..._on_Discrete_GPU_Market_Mercury_Research.html

Nvidia fought back with GTX 580, GTX 560 but it was not until Kepler that we saw huge market share shift in notebooks as Kepler was simple more efficient and there was no competition for mobile GK104. Recently that gap has massively widened as AMD Tonga was no answer for GM204. I do believe AMD has a R9 380X (Fiji XT) with 3072 sp and 4 GB HBM which is both cost and power optimized and which will also launch as a flagship notebook GPU with lower clocks and voltage binning.btw I am expecting 3 new GPUs -

Bermuda - 4096 sp, 8 GB HBM
Fiji - 3072 sp, 4 GB HBM
Trinidad - 1536 sp, 2/4GB GDDR5

I expect AMD to make strong share gains in notebook GPUs based on Fiji and Trinidad. We can continue to disagree but the eventual products will either prove or disprove our opinions.

 

[RussianSensation]

Until we see GTX 980 Ti specs and performance on launch day maybe you should reserve your AMD will be destroyed theme. :whiste:

This forum seems to have voted: $550 mid-range, $799 high-end, $999 flagship. AMD is not needed, hence should be destroyed as soon as possible because it's just staying in the way of progress and popularity contests. I can't wait for response from this forum at some point in the future when AMD will eventually have its own GPU generation that has superior perf/watt than NV. I wonder what key metric will be hyped then to justify buying NV, maybe price/performance or DP?

I expect AMD to make strong share gains in notebook GPUs based on Fiji and Trinidad.

Doubtful because R9 400 series are also re-brands of R9 300, and R9 500 series are re-brands of R9 400. No new GPUs ever made because re-brands FTW. When you'll see HBM on the box, it'll be mostly marketing but really HBM 1 and HBM 2 for AMD will just be re-branded GDDR5. "Greenland is the successor of Fiji, and we don’t think it will be radically new core." BTW, didn't you know AMD has permanently given up on mobile dGPUs until 2016? No new mobile dGPU strategy until 14nm. Starting next year, the entire AMD GPU division will focus on Nintendo NX though.

 

[AtenRa]

You are missing the point.

AMD makes smaller GPUs that come close to bigger NV GPU. The performance gap has been historically 15-20%.

The gap closed a great deal with Hawaii vs GK110, with the R290X beating Titan and eventually beating 780Ti.

Analyze it on a tech level. Die size, TDP, architecture, ram advantage. If AMD's latest 550mm2 die cannot beat GM200, it's a failure.

I dont actually remember if this has been confirmed, are we sure AMD will have a 550mm2 die ?

 

[showb1z]

I honestly don't care at all if the 390X beats the Titan. Buying the top SKU is a complete waste 95% of the time anyway. I really don't see what the miniscule €800+ market has to do with the vast majority of people who buy ~€200 cards.
Give me a 390 with performance somewhere between Titan and 980 for €400-450 and I'll buy it. I'm even willing to spend more than my usual ~€350 because of the 970 affair. Just release it already AMD.

And these 980Ti rumors seem like utter rubbish. Nvidia had to sell a gimped 970 but suddenly yields on the much bigger GM200 are perfect? No way.

 

[Silverforce11]

@RS & raghu78:
If you want to discount my speculation because its based on... speculation, then what are we talking about in these speculative threads?

Argue the point and whether it makes logical sense, rather than saying the point is rumor and we don't know if its true. If its the latter, nothing more to discuss.

Currently whats coming is a 2nd GM200 SKU, its miraculously supposed to be full GM200 with 6GB vram. If so, it should be faster than Titan X due to less TDP on the 1/2 vram, allowing higher core boost. It's also rumored to be $799.

Where does that leave room for AMD's 390X if its 20% slower than Titan X? How much do you think AMD could get for it if its that slow?

@AtenRa
Leaked last year I think, >500mm2 die size. I doubt its bigger than 600mm2 so 550 seems like a safe bet, in the ball-park.

 

[Silverforce11]

Sorry but you need to refresh your memory. AMD started making gains in both desktop and notebook GPUs starting with HD 4800 series and consolidated it with HD 5000 series.

Was a market-share success, but wasn't a commercial success because it was priced so low (5850/70 undercut NV so much!), it wasn't good at earning a profit. If you look at their revenue/profit during those times, they vary each quarter from a small loss, to break even, to a tiny profit.

In fact, AMD GPUs haven't been a commercial success, ever.

 

[boozzer]

I'm trying to point out that AMD hasn't ever been a commercial success when its slower and priced lower. It's a slow road to death.

it is just marketing bro, amd needs better, way better marketing. 80% of gamers/humans in general are sheep who would gobble up everything thrown at them without thought.

marketing is everything.

 

[AtenRa]

@AtenRa
Leaked last year I think, >500mm2 die size. I doubt its bigger than 600mm2 so 550 seems like a safe bet, in the ball-park.

I know about that leak but that doesnt confirm AMD will produce a 500-550mm2 die.

 

[Silverforce11]

I know about that leak but that doesnt confirm AMD will produce a 500-550mm2 die.

It's about as official as it gets because its from a firm that has worked with AMD to produce designs for fabs in in the past. They were working on two, a mid-range GPU and a massive >500mm2. The mid-range turn out to be Tonga.

 

[raghu78]

Was a market-share success, but wasn't a commercial success because it was priced so low (5850/70 undercut NV so much!), it wasn't good at earning a profit. If you look at their revenue/profit during those times, they vary each quarter from a small loss, to break even, to a tiny profit.

In fact, AMD GPUs haven't been a commercial success, ever.

rubbish. AMD's GPU division revenue and profit nos during the HD 5000 series and launch of HD 6000 series were

Q4 2009 - USD 427 million revenue USD 53 million profit
Q1 2010 - USD 409 million revenue USD 47 million profit
Q2 2010 - USD 440 million revenue USD 33 million profit
Q3 2010 - USD 390 million revenue USD 1 million profit
Q4 2010 - USD 424 million revenue USD 68 million profit

Profit as a % of revenue was decent. 12.4% , 11.4%, 7.5%, breakeven and 16%.

@RS & raghu78:
If you want to discount my speculation because its based on... speculation, then what are we talking about in these speculative threads?

Argue the point and whether it makes logical sense, rather than saying the point is rumor and we don't know if its true. If its the latter, nothing more to discuss.

Currently whats coming is a 2nd GM200 SKU, its miraculously supposed to be full GM200 with 6GB vram. If so, it should be faster than Titan X due to less TDP on the 1/2 vram, allowing higher core boost. It's also rumored to be $799.

Where does that leave room for AMD's 390X if its 20% slower than Titan X? How much do you think AMD could get for it if its that slow?

Since you are speculating on rumours let me tell you what I think will happen. I believe Nvidia will follow the same strategy as GK110. First a salvage SKU comes out with 2560 cc (5 GPC) and 384 bit memory bus at USD 700. It will land up around 10 - 15% slower than GTX Titan (so around 18 - 23% faster than GTX 980). My predictions for the R9 3xx series are consistent to chiphell leaks and based on what I eventually believe will be the chips config.

R9 390X WCE - 160%
R9 390X (Bermuda XT, 4096 sp) - 155%
Titan-X - 145%
R9 390 (Bermuda Pro, 3840 sp , 8 GB HBM)- 145%
GTX 980 Ti - 130%
R9 380X (Fiji XT, 3072 sp, 4 GB HBM) - 125%
R9 380 (Fiji Pro, 2816 sp, 4 GB HBM) - 115%
GTX 980 - 110%
R9 290X - 100%

So some time in Q4 2015 Nvidia will come out with newer SKUs like a overclocked full GM200 and a faster GTX 980 Ti with higher clocks(maybe Ultra edition with lesser cc disabled, say 2688 cc). There will be some fierce competition in perf and price which we as consumers will love.

 

[Silverforce11]

rubbish. AMD's GPU division revenue and profit nos during the HD 5000 series and launch of HD 6000 series were

Q4 2009 - USD 427 million revenue USD 53 million profit
Q1 2010 - USD 409 million revenue USD 47 million profit
Q2 2010 - USD 440 million revenue USD 33 million profit
Q3 2010 - USD 390 million revenue USD 1 million profit
Q4 2010 - USD 424 million revenue USD 68 million profit

Profit as a % of revenue was decent. 12.4% , 11.4%, 7.5%, breakeven and 16%.

Go back further to the 4800 series when GDDR5 started with AMD on the leading edge of memory.

Didn't help them much even when they had the tech edge. Loss, break even, paltry profits. Doesn't add up to much.

Also note, for a long time, the 5800 series had zero competition, no show from NV being late, and AMD didn't capitalize, they priced it so low it didn't generate a massive profit for them.

 

[raghu78]

Go back further to the 4800 series when GDDR5 started with AMD on the leading edge of memory.

Didn't help them much even when they had the tech edge. Loss, break even, paltry profits. Doesn't add up to much.

Also note, for a long time, the 5800 series had zero competition, no show from NV being late, and AMD didn't capitalize, they priced it so low it didn't generate a massive profit for them.

Did you even know what was happening in 2008 and 2009. the financial crisis and the following recession. Do you want me to show the Nvidia bleeding in those times.

http://www.nvidia.com/object/io_1234300989054.html

"For the fourth quarter of fiscal 2009, revenue was $481.1 million compared to $1.2 billion for the fourth quarter of fiscal 2008, a decrease of 60 percent. For the twelve months ended January 25, 2009, revenue was $3.4 billion compared to $4.1 billion for the twelve months ended January 27, 2008, a decrease of 16 percent.

NVIDIA's results for the fourth quarter of fiscal 2009, computed in accordance with U.S. generally accepted accounting principles (GAAP), included a net loss of $147.7 million, or $0.27 per share. Non-GAAP net loss for the fourth quarter of fiscal 2009, which excludes stock-based compensation charges, a benefit from insurance proceeds received, a non-recurring charge against operating expenses related to termination of a development agreement, a restructuring credit against operating expenses, and the associated tax impact of these items, was $94.4 million, or $0.18 per share."

http://www.nvidia.com/object/io_1241728875943.html

"During the first quarter of fiscal 2010, NVIDIA recorded a non-recurring charge of $140.2 million in connection with a previously announced cash tender offer to purchase employee stock options. This charge represents stock-based compensation expense associated with the stock options that were tendered, plus associated payroll taxes and professional fees.

NVIDIA's results for the first quarter of fiscal 2010, computed in accordance with U.S. generally accepted accounting principles (GAAP), included a net loss of $201.3 million, or $0.37 per share. Non-GAAP net loss for the first quarter of fiscal 2010, which excludes recurring stock-based compensation charges, the non-recurring charge related to the tender offer, and the associated tax impact of these items, was $46.7 million, or $0.09 per share."

http://www.nvidia.com/object/io_1249591520243.html

"NVIDIA Corporation today reported revenue of $776.5 million for the second quarter of fiscal 2010 ended July 26, 2009, down 13 percent from $892.7 million a year earlier, and up 17 percent from the previous quarter.
On a GAAP basis, the company recorded a net loss of $105.3 million, or $0.19 per share, compared with a net loss of $120.9 million, or $0.22 per share a year ago. Reflected in these results is a net charge of $119.1 million related to the weak die/packaging set used in certain previous-generation chips that the company had initially identified and provided for in its prior fiscal year. On a non-GAAP basis − excluding this item, and stock-based compensation as well as their associated tax impact − net income was $37.7 million, or $0.07 per diluted share, compared with $74.5 million, or $0.13 per diluted share, a year earlier"

2009 was a more difficult period for Nvidia as they had to fight an aggressive price war against AMD (remember HD 4890 and GTX 275 at USD 250) and the recession made it difficult for sales. It did not help that Nvidia also had to address the warranty problems due to GPU bump defects of products sold in 2007 and 2008. :whiste:

 

[3DVagabond]

Originally Posted by Silverforce11
Was a market-share success, but wasn't a commercial success because it was priced so low (5850/70 undercut NV so much!), it wasn't good at earning a profit. If you look at their revenue/profit during those times, they vary each quarter from a small loss, to break even, to a tiny profit.

In fact, AMD GPUs haven't been a commercial success, ever.

You posted this. ^^^

Originally Posted by raghu78
rubbish. AMD's GPU division revenue and profit nos during the HD 5000 series and launch of HD 6000 series were

Q4 2009 - USD 427 million revenue USD 53 million profit
Q1 2010 - USD 409 million revenue USD 47 million profit
Q2 2010 - USD 440 million revenue USD 33 million profit
Q3 2010 - USD 390 million revenue USD 1 million profit
Q4 2010 - USD 424 million revenue USD 68 million profit

Profit as a % of revenue was decent. 12.4% , 11.4%, 7.5%, breakeven and 16%

Go back further to the 4800 series when GDDR5 started with AMD on the leading edge of memory.

Didn't help them much even when they had the tech edge. Loss, break even, paltry profits. Doesn't add up to much.

Also note, for a long time, the 5800 series had zero competition, no show from NV being late, and AMD didn't capitalize, they priced it so low it didn't generate a massive profit for them.

Not as bad as you are making it out to be. No losses and profits that will make Lisa a hero if she can duplicate them. Let's not be too greedy and hope for a repeat.

 

[Silverforce11]

That's part of my point. Did AMD need to price war that aggressively with the 4800 series. I recall the GTX280 had a massive price cut as soon as the 4850/70 was released. It was completely un-necessary because the 4800 series was very competitive on all fronts, including perf & efficiency.

AMD price war never helped them since they needed a healthy profit to ensure further competition for latter generations.

Now this leads to the current & future situation. The market has changed a lot, NV is considered a premium brand. The 390X needs to be faster than Titan X else if its slower, it won't even sell well at $499 (due to the coming cut-down GM200). Selling a cutting edge large die with expensive HBM for so low isn't going to make them the required profits to stop the bleeding.

They simply cannot be slower, they need to reverse the negative brand image as well as dent into NV's premium status. The best way to do that is to pwn GM200 hard. Not be slower and selling for a lot less. That approach hasn't worked while NV's margins keeps getting better.

 

[3DVagabond]

That's part of my point. Did AMD need to price war that aggressively with the 4800 series. I recall the GTX280 had a massive price cut as soon as the 4850/70 was released. It was completely un-necessary because the 4800 series was very competitive on all fronts, including perf & efficiency.

AMD price war never helped them since they needed a healthy profit to ensure further competition for latter generations.

Now this leads to the current & future situation. The market has changed a lot, NV is considered a premium brand. The 390X needs to be faster than Titan X else if its slower, it won't even sell well at $499 (due to the coming cut-down GM200). Selling a cutting edge large die with expensive HBM for so low isn't going to make them the required profits to stop the bleeding.

They simply cannot be slower, they need to reverse the negative brand image as well as dent into NV's premium status. The best way to do that is to pwn GM200 hard. Not be slower and selling for a lot less. That approach hasn't worked while NV's margins keeps getting better.

I agree that this is likely a very expensive sku and they can't afford to offer it at bargain prices. They are badly in need of market share though and turning numbers, so who knows what they'll price it at.

Looking at the supposed specs, 4096 shaders, HBM, +550mm² it should be a real performer.

I still don't have an issue with their strategy with the 4000 and 5000 series. I think it worked as planned. It's Read's strategy that fell flat on it's face. I believe it's more to him that they have lost prestige with the brand than the 4000 and 5000 series. I don't think they did anything to hurt AMD's image.

 

[Silverforce11]

Not as bad as you are making it out to be. No losses and profits that will make Lisa a hero if she can duplicate them. Let's not be too greedy and hope for a repeat.

It was in regards to when "last time AMD had the memory edge" with GDDR5, that was from the 4800 series, which undercut NV massively, and continued to be low profit in the 5800 series. Overall if you add up the losses in 2008 to break even & small profits of 2009/10, it isn't much. AMD gained next to nothing (commercially, ie. profit, cash reserves) with their GDDR5 lead.

To be healthy, AMD needed their GPU division to pull off very strong profits to 1) recoup from the ATI buy and 2) subsidize the failing CPU dvision.

So I will stand by my statement that AMD hasn't been a commercial success by taking the road of 2nd place on performance and selling at a major discount compared to their competitor, not for CPU and not for GPUs.

The only option moving forward is to take the performance lead. 390X is actually their best chance to do so, ever, for the reasons which should be obvious: huge >500mm2 die (they always competed with smaller dies) combined with a new memory advantage.

 

[Skurge]

It was in regards to when "last time AMD had the memory edge" with GDDR5, that was from the 4800 series, which undercut NV massively, and continued to be low profit in the 5800 series. Overall if you add up the losses in 2008 to break even & small profits of 2009/10, it isn't much. AMD gained next to nothing (commercially, ie. profit, cash reserves) with their GDDR5 lead.

To be healthy, AMD needed their GPU division to pull off very strong profits to 1) recoup from the ATI buy and 2) subsidize the failing CPU dvision.

So I will stand by my statement that AMD hasn't been a commercial success by taking the road of 2nd place on performance and selling at a major discount compared to their competitor, not for CPU and not for GPUs.

The only option moving forward is to take the performance lead. 390X is actually their best chance to do so, ever, for the reasons which should be obvious: huge >500mm2 die (they always competed with smaller dies) combined with a new memory advantage.

Not sure why you are ignoring the evidence. 4800 and 5800 followed the 2900 and 3800 series. the GPU division increased market-share to almost even and was making a 10% profits DURING the recession when nVidia were making losses and losing marketshare.

I have no idea how you can call that strategy a failure. Those profits were huge at that time.