What will be AMD'S next Move?

Page 20 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

Abwx

Lifer
Apr 2, 2011
11,517
4,303
136
Can you elaborate? The card will not be using more power than it is during furmark (precisely why furmark throttles the card). So during luxmark the card is using less power (or identical) to furmark.

It looks like nvidia is simply doing what intel/AMD do for CPUs. Institute a power cap and throttle the card to fit in that cap (for reference).

On Furmark the card doesnt, and cant, provide its full throughput because it is power capped by the driver wich detect this benchmark, now if you think that the card use 150-160W in Luxmark just provide us the measurements that say so rather than doing unsubstancied claims.

On this graph we can see that at close to full throughput the 980 consume about 270-280W.

 
Last edited:

Wild Thing

Member
Apr 9, 2014
155
0
0
I expect R390 and R390XT will waltz in with a fast and innovative new architecture that will compete strongly with Maxwell part 1 as we see with GTX980/70.
GCN 2.0 is likely to retake the current NV lead and will do until Maxwell part 2 comes along next year.They will arrive with high bandwidth memory as well which could be a big help in making them fast,really fast.
If their liquid cooling systems from Asetek are an improvement on R295X2 then that's an effective way to match NV's more elegant power architecture.
Power usage becomes a lesser issue if it runs cool and fast.
 

AtenRa

Lifer
Feb 2, 2009
14,003
3,361
136
Can you elaborate? The card will not be using more power than it is during furmark (precisely why furmark throttles the card). So during luxmark the card is using less power (or identical) to furmark.

It looks like nvidia is simply doing what intel/AMD do for CPUs. Institute a power cap and throttle the card to fit in that cap (for reference).

The card throttles in Furmark because of the driver(it detects furmark and throttles to lower P state) not because of heat or workload or anything else.
 

CrazyElf

Member
May 28, 2013
88
21
81
Or more poorly optimized Ubisoft/EA/Activision "next gen" console ports with 2010-2012 graphics that still can't beat Crysis 3 but require us to get 4GB GPUs to hit 60 fps at 1080P.

I mean just imagine how demanding GTA V will be and how crappy it will look stock. :whiste:

I don't see that happening until 2016-2017 at which point 290/970 is a low-end card.

That's a serious problem. Far more likely we're going to get

Crysis 1 > 2/3, where game play became a lot more shallow and the games regressed in some ways.

The problem I see it is that there are so few companies that want to make a "deep" game with good gameplay, that also makes efforts to try to push ahead when it comes to graphics. What I'd love to see is something like the original Deus Ex, only pushing graphics the way Crysis did, or a really intensive RPG-type game.

Witcher 3 might be that "really intense" RPG.

As far as cheap monitors go, I'd say once they fall below <$1500 USD, we should start to see better adoption. It's got to be SST though and IPS (or at least a very good VA panel).

AMD's response should be strong as they must have foreseen the GTX 980 and GTX 970 when Nvidia launched GTX 750 Ti. I think AMD will have a response in Q1 2015 based on a next gen improved GCN 2.0 architecture.
...

Finally AMD is likely to introduce HBM with their next gen flagship GPUs. AMD has traditionally been the driver of graphics memory standards and is the first to market. AMD got GDDR5 out on HD 4870 around 2 years before Nvidia did with the GTX 480.

Considering it's likely that there will be no high power 20nm and we'll be stuck until 16nm FinFET becomes more affordable, it's likely that AMD will be the one as you note to jump to HBM first.

I also consider it likely in a few years that AMD will be the first likely to jump to 16nm (they have traditionally led when it comes to die shrinks).

Whether this can be translated into superior performance though is open for debate.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,595
136
Ryan Smith posted a message on a forum, (beyond3d?) read closely what he said.

Come on Abwx. I dont want to trawl all the forums. I cant find the post and you are the one knowing the power usage stuff.

Obviously there is something fishy here as the card uses less power under furmark than when gaming - lol . That shows NV have rigged the card, knowing the press would measure here. Just look at Ryans own testing. And yeaa its pathetic. But what else is there?
 
Last edited:

Enigmoid

Platinum Member
Sep 27, 2012
2,907
31
91
On Furmark the card doesnt, and cant, provide its full throughput because it is power capped by the driver wich detect this benchmark, now if you think that the card use 150-160W in Luxmark just provide us the measurements that say so rather than doing unsubstancied claims.

On this graph we can see that at close to full throughput the 980 consume about 270-280W.


I can't because I can't find any reviews.



Only thing I could find.

About the same as the 680.

BTW thats the wrong graph.
 

Enigmoid

Platinum Member
Sep 27, 2012
2,907
31
91
Ryan Smith posted a message on a forum, (beyond3d?) read closely what he said.

This is his quote.

http://forum.beyond3d.com/showthread.php?t=59531&page=92

For what it's worth, none of my compute benchmarks break TDP containment according to NVIDIA's drivers. You can definitely light up enough CUDA cores and push the card into throttling itself, but it's not violating TDP as far as I can tell. Power at the wall is consistent with FurMark.

Indicating that nvidia's drivers are throttling the card. Reference 980 will give you 165W of performance.

ABWX I don't understand you. You state that the card is using 280W on compute then quote someone who says it isn't.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,595
136
This is his quote.

http://forum.beyond3d.com/showthread.php?t=59531&page=92



Indicating that nvidia's drivers are throttling the card. Reference 980 will give you 165W of performance.

ABWX I don't understand you. You state that the card is using 280W on compute then quote someone who says it isn't.

With toms meassuring (not software), the OC 980 is using 280W on GPGPU - read at the wall so to speak?

As for Ryans remark (thanx for the link). There is two things here "according to NV drivers" - i read that as a cautionary sentence - , and the reference to furmark. As for furmark we know its rigged, because who would think furmark uses less power than a game? - so then what is the comparison to furmark worth?

Some testing shows the 980 getting neark 290x levels in some games:
http://www.hardware.fr/articles/928-7/consommation-efficacite-energetique.html

Obviously its not in all gaming situaitons the efficiency is held. And as furmark is heavvily throttled, its a difficult assessment.

I am also more interesting in the nature of the power fluctuations. Are they of voltage or amp of character?
 

nvgpu

Senior member
Sep 12, 2014
629
202
81
http://forum.beyond3d.com/showpost.php?p=1876415&postcount=2363

Tridam said:
You can run into this kind of extra power consumption in many games actually as Gigabyte boards keep their highest boost clock in every case (even in Furmark
). They increased significantly the TDP value in the bios, which is also the GPU Boost power target.

GTX 970 ref TDP : 150W
GTX 980 ref TDP : 180W
Gigabyte GTX 970 TDP : 250W
Gigabyte GTX 980 TDP : 300W

Obviously Gigabyte wanted to ensure the best possible performances in demanding cases but I'm not sure this perf improvement is worth the extra noise. Would have been better to provide this configuration as an optional one.

I also got those bios values for the Asus boards which are much more reasonable :

Asus GTX 970 TDP : 163.46W
Asus GTX 980 TDP : 195.14W

Please don't confuse what Gigabyte does with the power efficiency of Maxwell 2. Gigabyte cards clearly is running it at stupid high clocks and not caring about efficiency.
 

Enigmoid

Platinum Member
Sep 27, 2012
2,907
31
91
With toms meassuring (not software), the OC 980 is using 280W on GPGPU - read at the wall so to speak?

As for Ryans remark (thanx for the link). There is two things here "according to NV drivers" - i read that as a cautionary sentence - , and the reference to furmark. As for furmark we know its rigged, because who would think furmark uses less power than a game? - so then what is the comparison to furmark worth?

Some testing shows the 980 getting neark 290x levels in some games:
http://www.hardware.fr/articles/928-7/consommation-efficacite-energetique.html

Obviously its not in all gaming situaitons the efficiency is held. And as furmark is heavvily throttled, its a difficult assessment.

I am also more interesting in the nature of the power fluctuations. Are they of voltage or amp of character?

Toms is measuring 280W on the card only at the power input.

The reference 980 is pretty low power. The gigabyte models are overclocked and cannot be used for comparisons.

I do believe that the nvidia drivers are throttling the card to stay in tdp.

AT's clock speeds in game range from 1151 to 1227 a sizable difference. Likely the chip is adjusting clockspeeds to keep power and temps constant.
 

CakeMonster

Golden Member
Nov 22, 2012
1,492
653
136
Can't you have both? I'd like a factor oc'ed card (for the ease of it) with a cooler made to handle it but that still could idle at low power and throttle if necessary.
 

Abwx

Lifer
Apr 2, 2011
11,517
4,303
136
This is his quote.

http://forum.beyond3d.com/showthread.php?t=59531&page=92
Indicating that nvidia's drivers are throttling the card. Reference 980 will give you 165W of performance.

ABWX I don't understand you. You state that the card is using 280W on compute then quote someone who says it isn't.

It wont if there s the good drivers, in this case it will work exactly like in the THG graph posted above, peaks will reach 240/280W on the 970/980 but average power will be 150/180 or so, with drivers a la Gigabyte peak will somewhat increase but most importantly the average will be increased to the previous gen level, the reference cards numbers are here for marketing purposes only since Nvidia know that gamers are perfs hungry and that they will majoritarly buy the card that has the widest range even if they ll use it with Nvidia like settings.

Now on what interest us, the power management scheme, i m assuming that efficency is brought by power gating although i m not 100% sure but for the time i take this theory as a best case scenario, it s quite more complexe than a purely software approach since the latter will indeed be implemented in a heavy silicon solution.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,595
136
It wont if there s the good drivers, in this case it will work exactly like in the THG graph posted above, peaks will reach 240/280W on the 970/980 but average power will be 150/180 or so, with drivers a la Gigabyte peak will somewhat increase but most importantly the average will be increased to the previous gen level, the reference cards numbers are here for marketing purposes only since Nvidia know that gamers are perfs hungry and that they will majoritarly buy the card that has the widest range even if they ll use it with Nvidia like settings.

Now on what interest us, the power management scheme, i m assuming that efficency is brought by power gating although i m not 100% sure but for the time i take this theory as a best case scenario, it s quite more complexe than a purely software approach since the latter will indeed be implemented in a heavy silicon solution.

Is it possible to alter the driver to have lower peaks but higher average? I mean its demanding for the ps capacity and 970 could perhaps fit with one 6 pin if this tdp marketing showoff was lessened a little?
 

Cerb

Elite Member
Aug 26, 2000
17,484
33
86
Can't you have both? I'd like a factor oc'ed card (for the ease of it) with a cooler made to handle it but that still could idle at low power and throttle if necessary.
Having both of those things takes a lot of work. Intel, AMD, and nV all get better at it with each generation of part. AMD and nV themselves have good reason to set the stock specs near a power usage corner, where possible. Then, the other OEMs have to work around that. A new set of voltage regulators made for much higher current might be much less efficient under low loads, FI. There's no free lunch.
 

Cerb

Elite Member
Aug 26, 2000
17,484
33
86
It's curious that R9 290/290x price hasn't fallen at all. The retail price of R9 290 averages around $420, while the better performing GTX 970 is averaging around $360.

https://pcpartpicker.com/trends/video-card/#gpu.chipset.radeon-r9-290
It will happen. AMD will then keep the prices falling, and/or come out with new meh models. Then, they'll release a new competitive chip or two, and then the cycle will repeat. Even when they can, AMD doesn't want to release a new set of GPUs at the same time nVidia does, and Team Green feels the same. If either company has something really great, like Kepler and Maxwell have been, and the other isn't quite there, they'll look worse than if they just don't have anything new.

Fermi, which could run fast and hot, if allowed the voltage, but didn't in stock cards, due to AMD's huge efficiency lead (OEMs care about these things), is a good example. Likewise, back with the 8800 GTX, where they rebranded, re-spec'd, and re-priced, cards with that design a billion times, before finally coming out with new stuff. Today, AMD is the one on the butt end of it.
 

Ryan Smith

The New Boss
Staff member
Oct 22, 2005
537
117
116
www.anandtech.com

jpiniero

Lifer
Oct 1, 2010
15,108
5,665
136
Huh, I am a little surprised AMD hasn't cut prices further yet. The 290 is still going for $360 on Newegg. I can't imagine they are selling that many now.
 

Keysplayr

Elite Member
Jan 16, 2003
21,209
50
91
Huh, I am a little surprised AMD hasn't cut prices further yet. The 290 is still going for $360 on Newegg. I can't imagine they are selling that many now.

They will if Nvidia's partners run out of 970's at the rate they are selling.
 
Status
Not open for further replies.
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |