What will be AMD'S next Move?

[Kenmitch]

Bro if AMD was work in something than it would be leaked by now.They just cannot simply make a product in nights so it would take 5 to 6 months and even for Nvidia to counter.

Yes Titan Z price is far far ridiculous from Nvidia but they have learned and i am sure to say that GTX 970 is better than AMD any product that is available know expect R9 295X2 and it is much cheaper.

AMD has something in the works it's a given. The leaks or lack of leaks or rumors if you want to call them doesn't prove they do or don't.

Remember what happened with the untouchable Titan and the 290's launch?

Just for the record I'm eyeing the GTX 970 currently. I'm using the iGPU on my 4690k currently.

The constant bashing of AMD by the devoted green team members is somewhat of a turn off to me. You'd think that AMD pushing their favorite team to produce better products would be in their best interests.

 

[Abwx]

So u want to compare a $4xx gpu with a $330 gpu but still a $330 gpu wins by all means.

Does a 290 with a decent cooler cost 4xx $.?.

I would say that it s already at 350$ levels unless i missed something, here a 290 Tri X OC can be had for 354€ VAT 20% included, generaly our VAT priced cards are the same $ amounts for US users.

Besides, since you have abandonned the gaming experience as benchmark and are now focused on perf/watt irrespective of perfs i will point that a 290 with a good cooler will consume more on peaks than a regular 290 since it wont throttle at all but on average it will consume less, the difference in comsumption between a 970 and such a 290 is 50W at the card level on a demanding game, is that what you call a massive advantage.?.

 

[desprado]

Im sure AMD is thrilled. You are taking Toms power information out of context and building a story in your mind where AMD doesn't care.

that was really lol.

 

[desprado]

Does a 290 with a decent cooler cost 4xx $.?.

I would say that it s already at 350$ levels unless i missed something, here a 290 Tri X OC can be had for 354€ VAT 20% included, generaly our VAT priced cards are the same $ amounts for US users.

Besides, since you have abandonned the gaming experience as benchmark and are now focused on perf/watt irrespective of perfs i will point that a 290 with a good cooler will consume more on peaks than a regular 290 since it wont throttle at all but on average it will consume less, the difference in comsumption between a 970 and such a 290 is 50W at the card level on a demanding game, is that what you call a massive advantage.?.

Yes it cost more than $400 for a decent cooler where as EVGA which is the best brand is giving GTX 970 with a decent cooler at $339.

 

[Abwx]

You are taking Toms power information out of context and building a story in your mind where AMD doesn't care.

What is the context, that s nice to align words but does it even have a meaning or is it just an ad hominem attack since i you re just targeting me, but i see no argument since your "out of context" is just here to cover the vaccum of your argumentation or rather lack of...

 

[guskline]

Does a 290 with a decent cooler cost 4xx $.?.

I would say that it s already at 350$ levels unless i missed something, here a 290 Tri X OC can be had for 354€ VAT 20% included, generaly our VAT priced cards are the same $ amounts for US users.

Besides, since you have abandonned the gaming experience as benchmark and are now focused on perf/watt irrespective of perfs i will point that a 290 with a good cooler will consume more on peaks than a regular 290 since it wont throttle at all but on average it will consume less, the difference in comsumption between a 970 and such a 290 is 50W at the card level on a demanding game, is that what you call a massive advantage.?.

I was able to get a slightly used Sapphire R9 290 Tri-X for @$290 and buy a EK waterblock for $115 so for slightly more than $400 have a watercooled R9 290. BTW, the Tri-X is an outstanding air cooler.

 

[desprado]

AMD has something in the works it's a given. The leaks or lack of leaks or rumors if you want to call them doesn't prove they do or don't.

Remember what happened with the untouchable Titan and the 290's launch?

Just for the record I'm eyeing the GTX 970 currently. I'm using the iGPU on my 4690k currently.

The constant bashing of AMD by the devoted green team members is somewhat of a turn off to me. You'd think that AMD pushing their favorite team to produce better products would be in their best interests.

What we are saying is that many people here complained about nvidia pricing and that was the main reason why AMD survived but know that is also finished though so it is more a problem for AMD know since their product is more expensive and less better than Nvidia.

 

[f1sherman]

In the grand scheme of things, it really is Nvidia + AMD + Arm + everybody vs Intel. Having AMD around is beneficial to Nvidia and vice versa at this point. More particularly, it's everybody who uses TSC, GF, and Samsung vs Intel. Intel has a slight lead in the Fab business right now and a ever growing presence in the iGPU market. It's only a matter of time before the iGPU becomes a full blown GPU. You bet in the back of Nvidia and AMD's mind they know this. The day Intel begins their Tic Toc release of GPUs with a 1 1/2 node lead over them, it's eventually gg. AMD and Nvidia needs each other and Apple and Qualcomm to keep the pure Fab companies within a technological earshot of Intel. Actually, you can argue that the Iris Pro has already cost AMD and Nvidia revenue via the Mac Air contract. Remember, those things used to have a Geforce in it before the HD graphics and the Iris Pro. http://en.wikipedia.org/wiki/MacBook_Air

If AMD and Nvidia wants to be around in the long run, they need each other to gradually break into Intel's stranglehold in the server space and disrupt it's profit margin that's giving Intel the huge node need on everybody else. A lead that might very will one day be used in the GPU space.

AMD is certainly capable of energizing PC and gaming segment, which in turn helps Nvidia,
and I've heard JHH saying Nvidia needs and want strong AMD, but I'm not sure if that's an entirely honest sentiment.

I'm not saying that it's not. I really don't know/

 

[Genx87]

What is the context, that s nice to align words but does it even have a meaning or is it just an ad hominem attack since i you re just targeting me, but i see no argument since your "out of context" is just here to cover the vaccum of your argumentation or rather lack of...

I think it is pretty obvious. You are creating a story in your head based on Toms power consumption review. A review you clearly dont understand based on that thread. And from that misunderstanding are now claiming AMD is relieved to see Maxwell 2.0 isn't intrinsically advantaged in perf\watt over their current parts lol

Fighting the good fight is one thing. This is venturing into absurdity.

 

[Abwx]

Yes it cost more than $400 for a decent cooler where as EVGA which is the best brand is giving GTX 970 with a decent cooler at $339.

They could sell it 300$ instead, there s no reason that it has a higher manufacturing cost than the 970, the 290X on the other end will surely be placed somewhere between the 970 and the 980 in the waiting of their next GFX.

 

[Abwx]

I think it is pretty obvious. You are creating a story in your head based on Toms power consumption review. A review you clearly dont understand based on that thread. And from that misunderstanding are now claiming AMD is relieved to see Maxwell 2.0 isn't intrinsically advantaged in perf\watt over their current parts lol

Fighting the good fight is one thing. This is venturing into absurdity.

The perf/watt did improve in games according to Tom s review but also that if you extract the max throughput from thoses cards they will consume the same amount as the 780/780ti, so the perf/watt improvement is variable in function of the throughput, now if you dont agree explain us how you did understand this review s numbers, if i can return you the ad hominem i would say that its findings do not agree with what you would have hoped for, that is, that it s the same old architecture with power gatings and eventualy units with controled duty rates.

And i m waiting for your review of tom s review, btw, because set apart saying that i didnt understand it i dont read a lot of explanations in your posts, that s all ad hominem and to be frank i m tired of answering to such argumlentless "opinions"...

 

[Genx87]

Check the perfs before writing what is bs according to this review ;

http://www.xbitlabs.com/articles/graphics/display/amd-firepro-w9100.html

Firepro is far better, quality is as good and reliabilty is as good or even better since AMD GPUs were "tested" by litecoin miners, did you heard of problems..??..

And dont forget to give your opinion about the review and tell us by how much you estimate the Quadro advance..

Eh this is a comparison of AMDs newest vs the K5000. Which is a sub 2000 dollar Quadro based on the GK104 vs 3200+ for W9100. It also consumes over less than half the power of this Firepro(122 vs 275) and has 1/4th the memory(4GB vs 16GB).

Remember this is AMDs newest card released in Sept 2014 vs a card that came out in the Summer of 2012.

 

[USER8000]

In the last 12 years both AMD and Nvidia have gone through bad and good patches and are still here. All the doom and gloom threads for both sides,have been hilarious and happens at each GPU launch or price adjustment usually by gearheads.

When AMD is ahead certain people who own AMD/ATI cards or just hate Nvidia,predict Nvidia will die due to XYZ reasons,and likewise when Nvidia is ahead certain people who own Nvidia cards or hate AMD/ATI will say AMD/ATI will die. Its all just purchase justification E-PEEN on enthusiast forums in the end. Their chosen brand is the bestest and the other brand is the crappiest and should be burnt with fire,since it re-inforces their purchase choice as the best.

 

[positivedoppler]

AMD is certainly capable of energizing PC and gaming segment, which in turn helps Nvidia,
and I've heard JHH saying Nvidia needs and want strong AMD, but I'm not sure if that's an entirely honest sentiment.

I'm not saying that it's not. I really don't know/

Nvidia desperately need a strong AMD. Intel dumping of billions of dollars in Atom chips into the mobile market pretty bleed to death Nvidia's Tegra ambition. I don't think Intel would have been able to afford this if not for AMD's abysmal failure in CPU since bulldozer. Remember Intel once tried to buy Nvidia. I think at this point Intel has the graphic IPs needed to manufacturer GPUs regardless of Nvidia now. There should be a market demand for it too as Intel has the tools need now to address Low to Mid tier Graphics cards for those who want really low power consumption.

 

[SantaAna12]

What we are saying is that many people here complained about nvidia pricing and that was the main reason why AMD survived

With all due respect:
Are you being compensated in any way by Nvidia?

Warning issued for thread crapping.
-- stahlhart

 

[CrazyElf]

There seem to be a lot of "doomsayers" thinking that here's a 100% chance that AMD is dead in the water. This thread has been useful for one thing - figuring out who has a solid grasp of reality and who does not. I will keep it in mind in the future when I browse these forums. Let us consider the situation.

The GTX 980 is approximately 5% faster than the 780Ti and perhaps 10-12% faster than the 290X. That is not a huge performance gain. It has however gained perhaps ~40% in terms of power efficiency, which considering this is still the same process is a notable achievement. The higher the resolution, generally the more AMD closes the gap. I suppose that could translate into more overclocking headroom.

All in all, I would estimate that Maxwell has about a 60% lead in performance per watt in gaming (but not in compute) compared to the 290X. The 970, like the 290 compared to the 290X is by far a better value, although retailers have begun marking the prices upwards. Whether the price will go down in the future remains to be seen.

Scaling in 970/980 SLI does not appear to be as good as the 290/290X in Crossfire. You get about 65-70%, averaged across most games, compared to ~80% with CF. I wish that Nvidia would adopt a solution like AMD through PCI-E lanes. This is actually not a huge lead if you think about it.

The situation is certainly not as bad as say, the 8800 GTX vs the 2900XT. We have yet to see what AMD has to offer at this point. I will not judge until they release something. AMD has managed to make some impressive feats in the past, most notably their 4870 and 5870 cards.

The other interesting thing is whether or not a "large" Maxwell is coming >550mm^2. Very likely so. Personally that's what I plan to do - wait and see what AMD releases and see how it stands up against a big Maxwell. The 28nm process is quite mature, so we should be seeing a big die in the coming months - hopefully with shader, seeing that it is a mature process.

As far as the fans of either side - well very simply put, what the other side releases is not really important to you. You are never going to buy it anyways. It's only use is that it will lower prices for your side if it offers competitive performance. Remember - neither company is your friend, they just want your money.

 

[giantpandaman2]

There seem to be a lot of "doomsayers" thinking that here's a 100% chance that AMD is dead in the water. This thread has been useful for one thing - figuring out who has a solid grasp of reality and who does not. I will keep it in mind in the future when I browse these forums. Let us consider the situation.

The GTX 980 is approximately 5% faster than the 780Ti and perhaps 10-12% faster than the 290X. That is not a huge performance gain. It has however gained perhaps ~40% in terms of power efficiency, which considering this is still the same process is a notable achievement. The higher the resolution, generally the more AMD closes the gap. I suppose that could translate into more overclocking headroom.

All in all, I would estimate that Maxwell has about a 60% lead in performance per watt in gaming (but not in compute) compared to the 290X. The 970, like the 290 compared to the 290X is by far a better value, although retailers have begun marking the prices upwards. Whether the price will go down in the future remains to be seen.

Scaling in 970/980 SLI does not appear to be as good as the 290/290X in Crossfire. You get about 65-70%, averaged across most games, compared to ~80% with CF. I wish that Nvidia would adopt a solution like AMD through PCI-E lanes. This is actually not a huge lead if you think about it.

The situation is certainly not as bad as say, the 8800 GTX vs the 2900XT. We have yet to see what AMD has to offer at this point. I will not judge until they release something. AMD has managed to make some impressive feats in the past, most notably their 4870 and 5870 cards.

The other interesting thing is whether or not a "large" Maxwell is coming >550mm^2. Very likely so. Personally that's what I plan to do - wait and see what AMD releases and see how it stands up against a big Maxwell. The 28nm process is quite mature, so we should be seeing a big die in the coming months - hopefully with shader, seeing that it is a mature process.

As far as the fans of either side - well very simply put, what the other side releases is not really important to you. You are never going to buy it anyways. It's only use is that it will lower prices for your side if it offers competitive performance. Remember - neither company is your friend, they just want your money.

Pretty nice sum up. I just don't get why people get caught up in either camp. Team? Do people even know what team means?

In any case, I'm only posting to ask if anyone even knows when AMD is supposed to come up with their next die. Also, when are GPU companies going to be allowed to jump to the next process node?

A little roadmap so I could plan my upgrades would be nice. I've always wished Anandtech had a timeline feature where we could see CPU/GPU future upgrades as well as past upgrade milestones (along with soft launches vs. hard launches) as a service to their readers.

 

[Pariah]

VaporX runs about as cool and uses as much power as the 780 Ti… and it over clocks well.

http://www.hardocp.com/article/2014...290x_trix_oc_video_card_review/9#.VCBRllYtseM

Yes there was a disaster when the card was released. Yes the disaster was mitigated.

I don't think you understood what I meant. It doesn't matter what cooler you put on the card, it is still going to produce the same amount of heat. A better cooler only moves the heat away from the GPU more efficiently, and into your room more efficiently. The room with my computer in it is noticeably warmer than any other room in my house. It would be nice to close the temperature gap some by replacing major heat producing parts with more efficient ones. Not everyone wants to put up with the ill side effects of AMD's current similar to Nvidia's performance cards.

 

[realibrad]

I don't think you understood what I meant. It doesn't matter what cooler you put on the card, it is still going to produce the same amount of heat. A better cooler only moves the heat away from the GPU more efficiently, and into your room more efficiently. The room with my computer in it is noticeably warmer than any other room in my house. It would be nice to close the temperature gap some by replacing major heat producing parts with more efficient ones. Not everyone wants to put up with the ill side effects of AMD's current similar to Nvidia's performance cards.

Actually, that is not correct. If the cooler removes heat more effectively than another, it means that cooler can run the chip cooler. As heat builds in a chip, it becomes less efficient. As it becomes less efficient, it has to work harder to output the same performance. It does this, until it reaches a thermal limit. Once that thermal limit is hit, it slows down the chip.

So if a cooler in a gpu is efficient, it means the chip runs more efficient. This is a very basic way of explaining what happens, but I'm sure someone else can go deeper into the details.

 

[Pariah]

Actually, that is not correct. If the cooler removes heat more effectively than another, it means that cooler can run the chip cooler. As heat builds in a chip, it becomes less efficient. As it becomes less efficient, it has to work harder to output the same performance. It does this, until it reaches a thermal limit. Once that thermal limit is hit, it slows down the chip.

So if a cooler in a gpu is efficient, it means the chip runs more efficient. This is a very basic way of explaining what happens, but I'm sure someone else can go deeper into the details.

Yes, that is true in your physics book. It doesn't necessarily translate into the real world. From Anand's review of a custom cooled 290x:

"Under our gaming workload the reference 290 was already able to hit its maximum boost clock, so with the 290 Tri-X OC operating at a similar voltage and only a slightly higher clockspeed, we’re not seeing a meaningful increase in power consumption. This being despite the lower temperatures of the GPU, which would normally offer at least some savings due to reduced leakage. What this means is that custom coolers will not be able to do anything about the 290’s lesser weakness, which is its power consumption relative to the GTX 780."

 

[SimianR]

Yes, that is true in your physics book. It doesn't necessarily translate into the real world. From Anand's review of a custom cooled 290x:

"Under our gaming workload the reference 290 was already able to hit its maximum boost clock, so with the 290 Tri-X OC operating at a similar voltage and only a slightly higher clockspeed, we’re not seeing a meaningful increase in power consumption. This being despite the lower temperatures of the GPU, which would normally offer at least some savings due to reduced leakage. What this means is that custom coolers will not be able to do anything about the 290’s lesser weakness, which is its power consumption relative to the GTX 780."

Keep in mind it was the 290 (not the X) that they reviewed and that 290 Tri-X is a slightly overclocked card compared to the reference model. When I messed around with the Tri-X 290 briefly I found that underclocking to 950 and undervolting saw huge power savings. AMD was very aggressive with clocks on the 290/X to edge out the 770/780 and in the process really pushed the temps/power usage up. It would have been interesting to see a launch where perhaps they were slightly more conservative with the clockspeeds and thus power.

 

[realibrad]

Yes, that is true in your physics book. It doesn't necessarily translate into the real world. From Anand's review of a custom cooled 290x:

"Under our gaming workload the reference 290 was already able to hit its maximum boost clock, so with the 290 Tri-X OC operating at a similar voltage and only a slightly higher clockspeed, we’re not seeing a meaningful increase in power consumption. This being despite the lower temperatures of the GPU, which would normally offer at least some savings due to reduced leakage. What this means is that custom coolers will not be able to do anything about the 290’s lesser weakness, which is its power consumption relative to the GTX 780."

This graph shows differently.

http://www.techpowerup.com/reviews/Zotac/GeForce_GTX_480_Amp_Edition/27.html

 

[RussianSensation]

Abwx, you do realize that the consumer/gaming market still accounts for vastly the revenue for AMD right? I don't think your claims hold up when Maxwell IS very EE for games.

If I were an AMD shareholder, I would be crapping my pants in a panic and eagerly await their response.

No, you wouldn't because you'd actually analyse AMD a lot more closely today than assuming AMD today is similar to how AMD was 3-5 years ago.

In fact, what most people don't know is AMD's desktop dGPU division is the least valueable arm of AMD and is valued the least on a future going concern basis in how it translated to its forecasted stock price.

Only 7.2% of AMD's forecasted stock price is related to the AMD's discrete desktop GPU division.

Notebook CPUs = 20.07%
Server CPUs = 17.05%
Discrete notebook GPUs = 13.1%
Professional & Gaming Consoles = 11.3%
Desktop CPUs = 11.21%
Chipsets = 10.2%
Embedded CPUs = 9.92%
Discrete desktop GPUs = 7.2%
http://www.trefis.com/company?hm=AMD.trefis&from=search#/AMD/n-0856?from=sankey

So out of 7.2% of AMD's stock price, how much of that is related to GPUs in the $330+ price levels? Probably 30% at most or just 2.16% of AMD's total stock price if AMD were to lose 100% market share to NV in the $330+ level. AMD's notebook GPUs are nearly 2x as important for the firm as desktop GPUs. Where AMD needs to focus its efforts is mobile design wins even if it means giving up on competing with NV's desktop $500+ GPUs for 2-3 years until they have the finances and lower nodes available to them to resume competition at the high-end.

AMD still has their 265/270/270X/280/280X and these GPUs by far make the bulk of their revenues and profits for their desktop GPU division. As I said before AMD has at least 6-12 months to come up with a Maxwell competitor and even if GCN 2.0 isn't as good as Maxwell, as long as AMD gets more design wins in the mobile space, it could easily offset the losses they will have on the desktop if they do not have $400+ GCN 2.0 cards worth buying.

 

[f1sherman]

GM107, GM108 and G204 will devastate discrete mobile

why in your analysis you are disregarding Maxwell's impact on

• Discrete notebook GPUs = 13.1% of AMD value

and instead you are only concerned with $330+ desktop

 

[realibrad]

GM107, GM108 and G204 will devastate discrete mobile

why in your analysis you are disregarding Maxwell's impact on

• Discrete notebook GPUs = 13.1% of AMD value

and instead you are only concerned with $330+ desktop

In the context, gaming laptops are such a small market that it would also not make much of an impact.