[WCCFTECH]Possible AMD Radeon RX 490 Performance Numbers Show Up in DX12 AOTS Benchmark – On Par Wit

[frozentundra123456]

Perspective and semantics.

I would consider the 680/980/1080 mid range because they all had multiple products that perform much higher in their chip families. (no, the 1080Ti hasn't launched yet, but we all know it will)

I consider the 660/960/1060 "low end" for gaming and cards below that I don't consider. I'm not wealthy but I don't see much point in the X50 cards for desktop gaming. Might as well save money and buy a Xbox or PS4.

Well, it is pointless to continue to argue. The difference obviously is that I simply dont accept the definition of "high end" that it has to be a full die or the highest performance chip on the same die. High end is defined by the price and performance of the products currently available to the consumer. By your definition then, do you consider Fury X "high end" even though the 1080, and in most cases the 1070 outperform it? After all, it is the highest performing die from its chip family. You, and all the others who want to define the 1080 as "mid range", certainly are free to invent your own definition of the various classes of video cards, but by every conventional definition of "high end", the 1080 qualifies. You certainly are free to think the 1080 is too expensive, or wait for an assumed chip that will offer better value in the future, or buy a console, but that does not magically make a better performing chip currently available.

 

[advt.naveen]

You are comparing a theoretical mobile nano which never existed, and certainly never will, with a previous generation card that has been replaced. OK, that tells us a lot. The relevant comparison is between the most efficient current laptop chips. In fact, nVidia's architecture on 16nm is so efficient that they can essentially put a desktop chip into a laptop. So the comparison should be between a current mobile 1060/1070/1080 and whatever current AMD *laptop* chip offers comparable performance (if there is one).

AtenRa compared 980ti with nano , both are desktop chips.

 

[richaron]

Well, it is pointless to continue to argue. The difference obviously is that I simply dont accept the definition of "high end" that it has to be a full die or the highest performance chip on the same die. High end is defined by the price and performance of the products currently available to the consumer. By your definition then, do you consider Fury X "high end" even though the 1080, and in most cases the 1070 outperform it? After all, it is the highest performing die from its chip family. You, and all the others who want to define the 1080 as "mid range", certainly are free to invent your own definition of the various classes of video cards, but by every conventional definition of "high end", the 1080 qualifies. You certainly are free to think the 1080 is too expensive, or wait for an assumed chip that will offer better value in the future, or buy a console, but that does not magically make a better performing chip currently available.

You obviously don't understand the "invented" logic. What people do is compare all the chips in the current generation, and they call the fastest chip "high end". This top chip is not in the 1080. nV not currently offering the fastest for sale to consumers is a moot point, because it still exists.

Now if this chip (the one on topic) is the fastest chip AMD has made it will become AMD's "high end". See how that works? And it appears to have similar performance to the cards based on nVidia's less upper-mid chip in the 1070/80. But it's pretty obvious AMD are making 2 Vega chips and this is the lesser of them.

I'm still not 100% sure this isn't just dual Polaris tho, the performance and rumored power consumption seems right. I know the Aots screen didn't say mgpu, but does that mean it simply didn't recognise the product ID? If it is Vega it's a coincidence it's performing right around 1070/80 levels, but it wouldn't be the first time AMD cranked up the clocks (and down the efficiency) just to reach a performance target... I hope this is not the case again.

 

[AtenRa]

The last part here is not necessary. All it does it open yourself up to ridicule when your post indicates you don't know what you are talking about. The GTX980m and GTX 980 notebook are 2 unrelated products. Nvidia did a horrid job of naming them, but it is what it is. The GTX 980 notebook's memory runs at the exact same specs as the desktop version while the GPU base clock drops 62Mhz from 1126 to 1064. A 5.5% drop in clock rate with identical memory is not going to hurt performance very substantially.

No it doesnt

From NVIDIAs own site.


Desktop GTX 980
http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980/specifications

Memory Bandwidth (GB/sec) = 224

Mobile GTX 980m
http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-980m/specifications

Memory Bandwidth (GB/sec) = 160

Both on 254bit

 

[master_shake_]

People easily forget or chose not to talk about Fury Nano 175W TDP vs GTX 980Ti 250W TDP in 2016 games.

This card completely destroys GTX 980Ti in perf/watt at 1440p and especially in 4K.
Its amazing what this little card (28nm) can manage at 175W TDP and half the card size vs GTX980Ti.

i have nothing to say about the op or anything.

but 2016 i think was the death of dual gpu setups.

everyone could get by with itx/m-atx these days and not lose out on anything.... this sucks.

unless all you play is older games or total war.

 

[AtenRa]

You are comparing a theoretical mobile nano which never existed, and certainly never will, with a previous generation card that has been replaced. OK, that tells us a lot. The relevant comparison is between the most efficient current laptop chips. In fact, nVidia's architecture on 16nm is so efficient that they can essentially put a desktop chip into a laptop. So the comparison should be between a current mobile 1060/1070/1080 and whatever current AMD *laptop* chip offers comparable performance (if there is one).

As advt.naveen already said, my original comparison was Desktop Nano vs GTX 980ti which are both on the same gen same released dates etc etc.

Pariah for what ever reason compared a mobile GTX 980m to Desktop Nano, which is oranges to apples and when i replied with a Mobile Nano vs Mobile 980m you come here and complain for an unreleased product. Way to go.

 

[Vesku]

If this is GDDR5 or GDDR5X card it's either small Vega or a Polaris variant. Perhaps a Polaris design from TSMC fab, AMD did pay quite a bit to gain full Samsung/GF/TSMC production flexibility.

 

[Pariah]

No it doesnt

From NVIDIAs own site.


Desktop GTX 980
http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980/specifications

Memory Bandwidth (GB/sec) = 224

Mobile GTX 980m
http://www.geforce.com/hardware/notebook-gpus/geforce-gtx-980m/specifications

Memory Bandwidth (GB/sec) = 160

Both on 254bit

Why are you still talking about the GTX980M when you just quoted me saying the 980GTXM and the 980 GTX notebook are NOT the same product in the 2 sentences before the one you bolded?

THIS is the 980 GTX Notebook

This card is Nvidia's Nano for the Maxwell line. The GPU is the exact same one that is in the desktop 980 GTX, downclocked by 64 MHz (though OEM's can choose their base clock), with the exact same memory configuration and speed. The only difference is that unlike AMD with the Nano, Nvidia realized that no one was interested in paying a premium for a slightly slower and power optimized add in card version of a standard desktop gaming card. So instead they crammed it into a portable DR form factor where it provided unprecedented performance at its release.

Let's see how you side track this one instead of just admitting you were arguing about the wrong product.

 

[Glo.]

Why are you still talking about the GTX980M when you just quoted me saying the 980GTXM and the 980 GTX notebook are NOT the same product in the 2 sentences before the one you bolded?

THIS is the 980 GTX Notebook

This card is Nvidia's Nano for the Maxwell line. The GPU is the exact same one that is in the desktop 980 GTX, downclocked by 64 MHz (though OEM's can choose their base clock), with the exact same memory configuration and speed. The only difference is that unlike AMD with the Nano, Nvidia realized that no one was interested in paying a premium for a slightly slower and power optimized add in card version of a standard desktop gaming card. So instead they crammed it into a portable DR form factor where it provided unprecedented performance at its release.

Let's see how you side track this one instead of just admitting you were arguing about the wrong product.

Notebook GTX 980 has 150W TDP. 15W less than reference desktop GTX 980. Nano on the other hand has 175 TBP, and average 185W power consumption compared to 250W TBP, and 246W average power consumption on Fury X.

GTX 980 laptop can be considered as Nano, for Maxwell, but it still lacks on perf/watt.

End of off-top.

 

[Pariah]

Notebook GTX 980 has 150W TDP. 15W less than reference desktop GTX 980. Nano on the other hand has 175 TBP, and average 185W power consumption compared to 250W TBP, and 246W average power consumption on Fury X.

GTX 980 laptop can be considered as Nano, for Maxwell, but it still lacks on perf/watt.

End of off-top.

TDP is 125-150. Depends on OEM configuration. When you don't cherry pick only benchmarks that the Nano does well in, the efficiency drops significantly. The Nano's efficiency overall was not that special. THG tested it vs a 970 mini OC:

http://www.tomshardware.com/reviews/amd-radeon-r9-nano,4285-9.html

The Nano’s efficiency is better in one game, similar in two games and not far behind in all the other games compared to Nvidia's GeForce GTX 970 Mini (OC). The exception is Ashes of the Singularity, and the reason why was covered on the previous page.

So, overall a 970 was more efficient. The 980 mobile will certainly be more efficient than the 970 which would put it even further ahead of the Nano.

 

[Glo.]

TDP is 125-150. Depends on OEM configuration. When you don't cherry pick only benchmarks that the Nano does well in, the efficiency drops significantly. The Nano's efficiency overall was not that special. THG tested it vs a 970 mini OC:

http://www.tomshardware.com/reviews/amd-radeon-r9-nano,4285-9.html

So, overall a 970 was more efficient. The 980 mobile will certainly be more efficient than the 970 which would put it even further ahead of the Nano.

Why do you use very old benchmarks, when we have more accurate, current time data to calculate the efficiency of any GPU. Power consumption does not change over time. Performance with each release of drivers - can.

GTX 980M with 1536 CUDA cores have had 125W TDP. It was just a fraction slower than desktop GTX 970. Mobile version of GTX 980 was 150W TDP GPU. Only way you could push it down, was to push the clocks on the GPU and memory even further back.

I love when people on this forum cherry pick numbers for their agenda, and then moralize others about cherry picking numbers.

 

[AtenRa]

Why are you still talking about the GTX980M when you just quoted me saying the 980GTXM and the 980 GTX notebook are NOT the same product in the 2 sentences before the one you bolded?

THIS is the 980 GTX Notebook

This card is Nvidia's Nano for the Maxwell line. The GPU is the exact same one that is in the desktop 980 GTX, downclocked by 64 MHz (though OEM's can choose their base clock), with the exact same memory configuration and speed. The only difference is that unlike AMD with the Nano, Nvidia realized that no one was interested in paying a premium for a slightly slower and power optimized add in card version of a standard desktop gaming card. So instead they crammed it into a portable DR form factor where it provided unprecedented performance at its release.

Let's see how you side track this one instead of just admitting you were arguing about the wrong product.

Yes my bad i though you were talking about the GTX 980m because of the 120W TDP.

As for the Mobile GTX 980,

Why do you actually believe a mobile Nano will not be faster at 120W vs a 120W Mobile GTX 980 ???

If Nano has higher perf/watt at the 175W TDP vs 180W GTX 980 Desktop card, why the Mobile GTX 980 at 120W TDP will have higher perf/watt than a 120W Nano card ??

Especially when at lower TDPs, the HBM card will have an even greater advantage.

 

[Pariah]

Why do you use very old benchmarks, when we have more accurate, current time data to calculate the efficiency of any GPU. Power consumption does not change over time. Performance with each release of drivers - can.

GTX 980M with 1536 CUDA cores have had 125W TDP. It was just a fraction slower than desktop GTX 970. Mobile version of GTX 980 was 150W TDP GPU. Only way you could push it down, was to push the clocks on the GPU and memory even further back.

I love when people on this forum cherry pick numbers for their agenda, and then moralize others about cherry picking numbers.

One would assume that if you have more accurate benchmarks you would have posted them rather than just complaining that the ones posted don't support your opinion. The benchmarks posted above putting the Nano neck and neck and even ahead of a 980ti in multiple tests are by far the exception, not the norm. Are you going to try and claim that a Nano and 980ti are typically trading places?

The 980 notebook does not have an official TDP because it can be set by the OEM. This is reported by multiple sites. So claiming it is 150W and that's it is inaccurate.

 

[Glo.]

One would assume that if you have more accurate benchmarks you would have posted them rather than just complaining that the ones posted don't support your opinion. The benchmarks posted above putting the Nano neck and neck and even ahead of a 980ti in multiple tests are by far the exception, not the norm. Are you going to try and claim that a Nano and 980ti are typically trading places?

The 980 notebook does not have an official TDP because it can be set by the OEM. This is reported by multiple sites. So claiming it is 150W and that's it is inaccurate.

I don't because they are widely available. And I am not interested in proving it to you, because that will not change your mind, nor do I care about changing your mind about performance of Nano.

You used benchmarks from very very deep past as your proof for your agenda. Nobody cares about them in 2016, when we have benchmarks that show that Nano is performing much better, than in 2015. It is only your problem about changing your mind.

Yes, GTX 980 MXM has variable TDP. But it comes with a cost. GTX 980 nominal is 150W GPU. And even at that it will be much slower than Nano. That is what will benchmarks show, regardless of how you want to argue about it.

 

[Pariah]

Yes my bad i though you were talking about the GTX 980m because of the 120W TDP.

As for the Mobile GTX 980,



If Nano has higher perf/watt at the 175W TDP vs 180W GTX 980 Desktop card, why the Mobile GTX 980 at 120W TDP will have higher perf/watt than a 120W Nano card ??

Especially when at lower TDPs, the HBM card will have an even greater advantage.

I'm not really interested in arguing this topic any more. The choice to pick the 980ti as the Nvidia representative for efficiency against the Nano IMO was a poor one as the primary focus for the engineering teams were not the same. The 980 notebook is a more valid comparison in that regard. Whether or not one or the other eeks out a slight advantage I don't care enough to continue this conversation. When comparing like targetted products the efficiencies in this particular case are close enough to not give either side any real bragging advantage.

Tip of the cap for admitting the wrong card mistake.

 

[esquared]

Stop with the thread derailing. This is not a GTX 980 thread.

Post to the topic. Radeon. RX 490.


esquared
Anandtech Forum Director

 

[tajoh111]

People easily forget or chose not to talk about Fury Nano 175W TDP vs GTX 980Ti 250W TDP in 2016 games.

This card completely destroys GTX 980Ti in perf/watt at 1440p and especially in 4K.
Its amazing what this little card (28nm) can manage at 175W TDP and half the card size vs GTX980Ti.

Nope.

http://www.hardware.fr/articles/942-23/fiji-vs-gm200-185w.html

When given the same power limit and usage, the gtx 980 ti outperforms the nano.

Underclock and undervolting goes both ways, meaning there is savings in maxwell as well. Same goes for overclocking.

https://www.techpowerup.com/reviews/AMD/R9_Fury_X_Overvoltage/2.html

Fury x needs overvolting to do any sort of significant overclocking and power consumption jumps up like crazy. 15.7% increase in frequency and you get a 66% increase in power consumption.

While nvidia cards can be overclocked 25% and they only consume 15 watts more than stock.

https://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_XtremeGaming/23.html

https://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_XtremeGaming/21.html

http://www.anandtech.com/show/9621/the-amd-radeon-r9-nano-review/18

"Otherwise on a broader competitive basis, while AMD is enjoying an improvement in energy efficiency they still must contend with NVIDIA, and here is where things get murky. Relative to the rest of their lineup the R9 Nano’s energy efficiency is fantastic. However relative to the energy efficiency of NVIDIA’s lineup, AMD’s gains mostly serve to close the gap that already existed. While admittedly only painting the broadest of strokes here, R9 Nano demonstrates slightly better performance than GTX 980 –around 5% at 2560x1440 – for a similar increase in power consumption. Which is to say that AMD has closed the gap, but they don’t have any real lead."

Nano is more a product of marketing and rebranding what most people knew. Undervolting and underclocking can save significant amounts of power no matter what the architecture.

The fact that the gtx 980 ti underclocks and overclocks well, just means it has fury nano or fiji generally beat unless you need the form factor.

Add the 650 price tag of nano, and it was a project doomed from the start. Even without competition because of its form factor, the fury nano dropped 100-150 dollars in price in two months. A sign of poor marketing analysis and bad execution on AMD's part.



Right above your post I said to keep this on-topic.
This is not a GTX980 thread. Its an RX490 thread.


esquared
Anandtech Forum Director

 

[Glo.]

Based on what we know about 14 nm GloFo process, porting 4096 GCN core chip from 28 nm to 14 nm, with 185W TDP it would mean the chip can have 1250-1300 MHz core clock, at that TDP.

50% higher performance at the same thermal envelope. Nano has 850 MHz core average frequency. 10 TFLOPs of compute power.

However, we do not know exactly how would new GPU behave on smaller node. We have seen previously efficiency related patents, and already discussed them. One of the things that has been more and more apparent about the patents was that Vega architecture may be configured to achieve higher clocks.

Second thing about the patents is that to get the best of it in GCN architecture would require redesing of scheduling, and redesigning the shader engine layout. Fiji have had 4 Shader Engines. Vega 10 with 64 CU's can have 8, for better efficiency(power gating unused shader engines), and at the same time higher overall throughput.

What can we expect? Well if the rumors are true, we can expect 1.5 GHz core clock on 4096 GCN core part, and 12 TFLOPs of compute power. That alone would put it directly on the Titan X pascal level. But we do not know how it would behave with the improvements of next generation architecture.

My predictions for Vega 10 and 11:

Vega 11 - 300-350 mm2 die size, 3072 GCN cores, 1.5 GHz, 8 GB of HBM2, 512 GB/s, 150-175W TDP.
Vega 10 - 400-450 mm2 die size, 4096 GCN cores, 1.5 GHz, 16 GB of HBM2, 512 GB/s, 225W TDP.

 

[Dave2150]

I think it's wise that we all remember Vega might be a June 2017 product. It might actually launch after NVIDIA's Volta.

If it were releasing soon, we'd have heard about it by now. There'd be actual leaks, whereas there is absolutely nothing concrete at all.

 

[Arachnotronic]

I think it's wise that we all remember Vega might be a June 2017 product. It might actually launch after NVIDIA's Volta.

If it were releasing soon, we'd have heard about it by now. There'd be actual leaks, whereas there is absolutely nothing concrete at all.

While I would love me some Volta action, I think we will see Vega first.

 

[Dave2150]

While I would love me some Volta action, I think we will see Vega first.

I hope so

 

[frozentundra123456]

You obviously don't understand the "invented" logic. What people do is compare all the chips in the current generation, and they call the fastest chip "high end". This top chip is not in the 1080. nV not currently offering the fastest for sale to consumers is a moot point, because it still exists.

Now if this chip (the one on topic) is the fastest chip AMD has made it will become AMD's "high end". See how that works? And it appears to have similar performance to the cards based on nVidia's less upper-mid chip in the 1070/80. But it's pretty obvious AMD are making 2 Vega chips and this is the lesser of them.

I'm still not 100% sure this isn't just dual Polaris tho, the performance and rumored power consumption seems right. I know the Aots screen didn't say mgpu, but does that mean it simply didn't recognise the product ID? If it is Vega it's a coincidence it's performing right around 1070/80 levels, but it wouldn't be the first time AMD cranked up the clocks (and down the efficiency) just to reach a performance target... I hope this is not the case again.

I understand your "logic" perfectly. I just disagree with you.

 

[Pariah]

Vega will make it to market before Volta is announced. Officially unveiling the volta architecture in the first half of 2017, even without any specific product announcement would prevent Nvidia from releasing any more Pascal products. Too much money was spent on the development of Pascal for there to not be a refresh. The odds of Nvidia moving to Volta without ever offering a "Ti" Pascal to fill the gap between the $1200 Titan and $600 1080GTX are pretty much zero. That would leave way too much money on the table.

There will not be a 16GB HBM2 Vega product either. The price difference between the 12GB Nvidia P100 and 16GB is $1475. That's $1475 for 4GB of HBM2. Now, of course it doesn't cost anywhere near that much, but they would be hard pressed to charge that much for an extra 4GB of GDDR5. HBM2 is much more expensive than GDDR5, how is AMD going to make any money trying to sell a 16GB card for under $1000 while Nvidia is charging 50% more for just 4GB of the RAM and nothing else? AMD is not in a position where they can inflate product prices by adding really expensive superfluous features that won't improve performance like useless extra RAM.

 

[OatisCampbell]

Well, it is pointless to continue to argue. The difference obviously is that I simply dont accept the definition of "high end" that it has to be a full die or the highest performance chip on the same die. High end is defined by the price and performance of the products currently available to the consumer. By your definition then, do you consider Fury X "high end" even though the 1080, and in most cases the 1070 outperform it? After all, it is the highest performing die from its chip family. You, and all the others who want to define the 1080 as "mid range", certainly are free to invent your own definition of the various classes of video cards, but by every conventional definition of "high end", the 1080 qualifies. You certainly are free to think the 1080 is too expensive, or wait for an assumed chip that will offer better value in the future, or buy a console, but that does not magically make a better performing chip currently available.

The Fury X I considered "high end" when it launched because it was within 4-6% of the performance of the top card of the time, the Titan X. (at 4K, top standard resolution)

https://www.techpowerup.com/reviews/AMD/R9_Fury_X/31.html

Now I consider the Fury X midrange because it's 42% slower than the current top card at 4K:

https://www.techpowerup.com/reviews/NVIDIA/Titan_X_Pascal/24.html

I only think about consumer cards, but the 1080 is about to become the 3rd fastest card out there when NV launches their 1080Ti, and maybe the fourth if AMD can beat it with Vega.

3rd or 4th is high midrange to me, but I'll give you that 1080s are technically high end until the Ti, and possibly Vega, launch. (not that it matters what I think)

 

[Triloby]

I'm not going to get my hopes up with the 490. Considering how much of a disappointment Polaris and the Fury cards were out of the gate, I highly doubt that Vega will be as impressive as AMD makes it out to be. Even if the 490 manages to compete with the GTX 1080, it probably won't stand a chance against the incoming 1080 Ti.

All that matters to AMD now is Zen. If they can't deliver on that, then Vega will be even more meaningless in the grand scheme of things (even if Vega does manage to go above expectations).