Overclock.net banner

[DSOgaming] AMD confirms that it will showcase its new Radeon RX Vega gaming/consumer graphics card at Computex

23K views 309 replies 93 participants last post by  TheBlademaster01 
#1 ·
#2 ·
#4 ·
Quote:
Originally Posted by xioros View Post

If it's a pascal killer, it's late to the party. It better be a volta competitor (although we all know it won't be)
Still having hard time to understand, why everybody needs Vega above 1080Ti or Titan Xp...
As long as it matches at least 1070 with reasonable price, it will be fine.
AMD/RTG has different budget than nVidia, and it seems to me that everybody is just blind to that fact.
rolleyes.gif
 
#5 ·
Quote:
Originally Posted by BobiBolivia View Post

Still having hard time to understand, why everybody needs Vega above 1080Ti or Titan Xp...
As long as it matches at least 1070 with reasonable price, it will be fine.
AMD/RTG has different budget than nVidia, and it seems to me that everybody is just blind to that fact.
rolleyes.gif
Because 1070 performance is only ~10% faster than a fury x?
 
#7 ·
Quote:
Originally Posted by BobiBolivia View Post

Still having hard time to understand, why everybody needs Vega above 1080Ti or Titan Xp...
As long as it matches at least 1070 with reasonable price, it will be fine.
AMD/RTG has different budget than nVidia, and it seems to me that everybody is just blind to that fact.
rolleyes.gif
Agreed. I bet the 1070 has the highest sale numbers out of the pascal cards (in all formats).
Its a powerful card that does not break the bank.
As proven by the 480, its not the most expensive cards that rank in the $$$
I personally can only consider Volta if it is at around 1080Ti levels as i will need a decent jump over my 1070, but if its not, that only means i loose out, not that the entire brand/rang is a failure...
 
#8 ·
Quote:
Originally Posted by BobiBolivia View Post

Still having hard time to understand, why everybody needs Vega above 1080Ti or Titan Xp...
As long as it matches at least 1070 with reasonable price, it will be fine.
AMD/RTG has different budget than nVidia, and it seems to me that everybody is just blind to that fact.
rolleyes.gif
It doesn't need to be, it's just stuff people say in order to maintain an AMD sucks stance. Even if it's good, it still sucks because Nvidia is launching a new lineup of cards next year that will beat it.

I have a waterblock on the way for a 1080ti that I am picking up in a trade and I still plan on getting a Vega card when they come out. It doesn't have to be better than any particular card for people to buy it.
 
#9 ·
Quote:
Originally Posted by BobiBolivia View Post

Still having hard time to understand, why everybody needs Vega above 1080Ti or Titan Xp...
As long as it matches at least 1070 with reasonable price, it will be fine.
AMD/RTG has different budget than nVidia, and it seems to me that everybody is just blind to that fact.
rolleyes.gif
Well considering the fury x is already within a few % of the 1070... I think it is a pretty sure guarantee Vega will be faster than the 1070 lol. If i was to make a bet, I would say that the top Vega product is almost exactly 1.5x as fast as the fury x.

Clock speeds alone should be like 1.5x the Fury X, scaling wouldn't be perfect so i think that 10-15% of diminishing returns will be replaced with arch improvements to get to 1.5x. That should put it somewhere in between the 1080 and the 1080ti, perhaps leaning a bit more towards the 1080ti.

Enthusiasts that prefer AMD/use freesync are just a bit disappointed they haven't had anything offered to them at the top end of graphics performance for a while now.
 
#10 ·
I think if they price this well it will do ok even if it doesn't beat the current Nvidia offerings. Its going to have to be priced very competitively.

With such a late release its not looking good.
 
#11 ·
#12 ·
Quote:
Originally Posted by BobiBolivia View Post

Still having hard time to understand, why everybody needs Vega above 1080Ti or Titan Xp...
As long as it matches at least 1070 with reasonable price, it will be fine.
AMD/RTG has different budget than nVidia, and it seems to me that everybody is just blind to that fact.
rolleyes.gif
I would like it to be a little faster than the 1070 since I really don't want to do a side grade. I really want Vega so I plan to get one IF the performance is worth my time and it is priced correctly.
 
#13 ·
AMD already has experience with 14nm GPUs, so Vega should have similar IPC to Pascal. If you could scale a Fury X to 1600mhz and give it an extra 4gb HBM, it would be in GTX 1080 territory. I think 1080ti performance is within reach.
 
#15 ·
Can anyone find out what time AMD are presenting at computex on the 31st?
I cannot find an actual time, just that AMD will be presenting on the 31st....
Taipei is GMT + 8:00 so that's 13 hours ahead of me (EST). Hope we get to wake up to some EPIC news.
 
#16 ·
Quote:
Originally Posted by Ultracarpet View Post

Well considering the fury x is already within a few % of the 1070...
Too bad it is not that easy to compare.

Fiji was a humongous chip with 596mm² in 28nm and had 275W TDP in the strongest version.The GTX 1070 is 16nm FinFet has around 314mm² die size and is listed for 150W TDP. Keep in mind AMd already had HBM used in Fiji cards, so there is no further power/heat advantage to gain.

The difference in manufacturing costs is pretty bad. AMD would have to improve the Vega chip quite dramatically. A simple shrink and some adjustments won't do. The costs and availability of HBM2 are another factor to count in.

Quote:
Originally Posted by Ultracarpet View Post

I think it is a pretty sure guarantee Vega will be faster than the 1070 lol. If i was to make a bet, I would say that the top Vega product is almost exactly 1.5x as fast as the fury x.
The question is which Vega will be faster than the 1070. Some cut down version will be competitors to the 1070 too, yes. But the upper versions can not be simple multiplied like this. The bigger the chip / performance class is, the more difficult it gets to get all the shaders and transistors utilized. The tech doesn't scale infinitely. At some point you have dimishing returns.

Quote:
Originally Posted by Ultracarpet View Post

Clock speeds alone should be like 1.5x the Fury X, scaling wouldn't be perfect so i think that 10-15% of diminishing returns will be replaced with arch improvements to get to 1.5x. That should put it somewhere in between the 1080 and the 1080ti, perhaps leaning a bit more towards the 1080ti
.

The clock speeds alone say nothing if there is no information how much the architecture itself scales. Your 10-15% dimishing returns may occure every +60 MHz form Fiji. At the same time the power requirements and heat output will rise similar to what AMd showed with Polaris 2.0 in the RX580. I personally say the outcome will highly be dependent on the scenario. Like Fiji had before, Vega will show better results i.e. higher resolutions?

Quote:
Originally Posted by Ultracarpet View Post

Enthusiasts that prefer AMD/use freesync are just a bit disappointed they haven't had anything offered to them at the top end of graphics performance for a while now.
Not entirely correct. Everybody above the $300 region has no option to go, if AMD is the prefered brand. This fact and the long waiting period is starting to annoy peopel to no end. In case AMD expands this another few months, some might be no longer patient and buy whatever is already in their budget... or just don't buy anything and wait for the next generation of GPUs.

Quote:
Originally Posted by 0451 View Post

AMD already has experience with 14nm GPUs, so Vega should have similar IPC to Pascal. If you could scale a Fury X to 1600mhz and give it an extra 4gb HBM, it would be in GTX 1080 territory. I think 1080ti performance is within reach.
IPC has not to do with the process node. It can only alter it slightly. IPC is mostly dependent on the used architecture. GCN as it was until Polaris does not scale very well. AMD had trouble utilizing all their shaders and had to resort to tricks like Async Compute. Their HBCC is the next step to get the 4096 shaders under control.

No matter where the Vega cards finally land, their price and the availability in the next two months will be a very deciding factor. Fiji failed at both, so let's hope AMD gets the HBM2 shortage under control very soon.
 
#17 ·
Quote:
Originally Posted by Hardware Hoshi View Post

Too bad it is not that easy to compare.

Fiji was a humongous chip with 596mm² in 28nm and had 275W TDP in the strongest version.The GTX 1070 is 16nm FinFet has around 314mm² die size and is listed for 150W TDP. Keep in mind AMd already had HBM used in Fiji cards, so there is no further power/heat advantage to gain.

The difference in manufacturing costs is pretty bad. AMD would have to improve the Vega chip quite dramatically. A simple shrink and some adjustments won't do. The costs and availability of HBM2 are another factor to count in.
The question is which Vega will be faster than the 1070. Some cut down version will be competitors to the 1070 too, yes. But the upper versions can not be simple multiplied like this. The bigger the chip / performance class is, the more difficult it gets to get all the shaders and transistors utilized. The tech doesn't scale infinitely. At some point you have dimishing returns.
.

The clock speeds alone say nothing if there is no information how much the architecture itself scales. Your 10-15% dimishing returns may occure every +60 MHz form Fiji. At the same time the power requirements and heat output will rise similar to what AMd showed with Polaris 2.0 in the RX580. I personally say the outcome will highly be dependent on the scenario. Like Fiji had before, Vega will show better results i.e. higher resolutions?
Not entirely correct. Everybody above the $300 region has no option to go, if AMD is the prefered brand. This fact and the long waiting period is starting to annoy peopel to no end. In case AMD expands this another few months, some might be no longer patient and buy whatever is already in their budget... or just don't buy anything and wait for the next generation of GPUs.
IPC has not to do with the process node. It can only alter it slightly. IPC is mostly dependent on the used architecture. GCN as it was until Polaris does not scale very well. AMD had trouble utilizing all their shaders and had to resort to tricks like Async Compute. Their HBCC is the next step to get the 4096 shaders under control.

No matter where the Vega cards finally land, their price and the availability in the next two months will be a very deciding factor. Fiji failed at both, so let's hope AMD gets the HBM2 shortage under control very soon.
Not sure if you read anything in my post lol.
 
#18 ·
Quote:
Originally Posted by SystemTech View Post

Can anyone find out what time AMD are presenting at computex on the 31st?
I cannot find an actual time, just that AMD will be presenting on the 31st....
Taipei is GMT + 8:00 so that's 13 hours ahead of me (EST). Hope we get to wake up to some EPIC news.
Statement from AMD:

Computex 2017 is fast approaching so we wanted to share a save-the-date for the AMD press conference, scheduled for May 31st from 10 a.m. - 11 a.m. Hosted by AMD CEO, Dr. Lisa Su and other key AMD executives, you will have the opportunity to hear more about the latest products and leading-edge technologies coming from AMD in 2017. The past year has seen AMD bringing innovation and competition back to the high-performance desktop market with the release of Ryzen™ processors and we look forward to providing new details on 2017 products and the ecosystems, both OEM and channel, that will support them.

13 hrs behind? That make 9pm May 30th for you?
 
#19 ·
Quote:
Originally Posted by WiggleToes View Post

Statement from AMD:

Computex 2017 is fast approaching so we wanted to share a save-the-date for the AMD press conference, scheduled for May 31st from 10 a.m. - 11 a.m. Hosted by AMD CEO, Dr. Lisa Su and other key AMD executives, you will have the opportunity to hear more about the latest products and leading-edge technologies coming from AMD in 2017. The past year has seen AMD bringing innovation and competition back to the high-performance desktop market with the release of Ryzen™ processors and we look forward to providing new details on 2017 products and the ecosystems, both OEM and channel, that will support them.

13 hrs behind? That make 9pm May 30th for you?
Good find thank you good sir.
Whoop whoop, only 11 days to go hehe.
 
#20 ·
Quote:
Originally Posted by Ultracarpet View Post

Not sure if you read anything in my post lol.
I did and quoted every part of your post. You make it sound way to easy. Every bit of improvement has to be developed, tested and planned. Nobody in the IT-industry gets X amount of performance out of nowhere. The results entirely depend for what AMD develops the chip for.

AMD is said to use the same chip design for both consumer and professional line-ups again. Since there is so much compute ballast in there not useful for gaming, the total gamign performance might suffer from this. Even if theoretically possible, your +50% improvements might not be met in any situation.

Quote:
Originally Posted by The EX1 View Post

All I get from Raja's reply is that they are still behind on Vega. When they, AMD, start talking about how difficult their job is, that is a red flag for meeting release date expectations.
So it seems. I got the opinion the Ryzen launch came at the wrong time for the RTG group. Their funds were cut because of this and from 2010 on alot of their former personal got fired. The AMD R&D budget was at an all time low and quarter after quarter only losses came in financially. The fixation on HBM-based technology put them quite in a bind.

The RTG engineers must have sweated blood and dirt to accomplish anything. Hard work indeed, but so is everything in a working environment. AMD won't get any browny points for what they do. It's their god damn job a boss would say.
 
#22 ·
Quote:
Originally Posted by Hardware Hoshi View Post

I did and quoted every part of your post. You make it sound way to easy. Every bit of improvement has to be developed, tested and planned. Nobody in the IT-industry gets X amount of performance out of nowhere. The results entirely depend for what AMD develops the chip for.

AMD is said to use the same chip design for both consumer and professional line-ups again. Since there is so much compute ballast in there not useful for gaming, the total gamign performance might suffer from this. Even if theoretically possible, your +50% improvements might not be met in any situation.
Considering the only difference in the 8.5tflops of the furyx to the 13tflops of vega is a clockspeed improvement (which is roughly 1.5x) I don't really understand what your point is. Theoretically, the GPU should straight up just be 1.5x more powerful, and I think in the end, the improvements to the arch will counter-act the diminishing returns of a straight up increased clock speed.
 
#23 ·
Quote:
Originally Posted by Ultracarpet View Post

Considering the only difference in the 8.5tflops of the furyx to the 13tflops of vega is a clockspeed improvement (which is roughly 1.5x) I don't really understand what your point is. Theoretically, the GPU should straight up just be 1.5x more powerful, and I think in the end, the improvements to the arch will counter-act the diminishing returns of a straight up increased clock speed.
Everything is being clocked up... the ROPs, the TMUs, the Geometry Engines, the Cache etc.

So yeah... it is a straight up 1.5x more powerful GPU with tweaks which should lower the cycles per instruction on the Geometry, ROP, Compute fronts.

In other words... you're right and have no explaining to do... that other guy Hardware Hoshi... well... you can ignore what he says
biggrin.gif


PS. Raja stated that the gaming Vega tweaks specific to RX Vega not on the Vega frontier cards as well.
 
#24 ·
I don't know why some people make it sound like 1.5x of Fury X performance is somewhat impossible to reach lol. It is not that high of a goal.
 
#25 ·
Quote:
Originally Posted by Kpjoslee View Post

I don't know why some people make it sound like 1.5x of Fury X performance is somewhat impossible to reach lol. It is not that high of a goal.
It's what I'm expecting pretty much, anything above that is a pleasant surprise
tongue.gif

Quote:
Originally Posted by Mahigan View Post

Everything is being clocked up... the ROPs, the TMUs, the Geometry Engines, the Cache etc.

So yeah... it is a straight up 1.5x more powerful GPU with tweaks which should lower the cycles per instruction on the Geometry, ROP, Compute fronts.

In other words... you're right and have no explaining to do... that other guy Hardware Hoshi... well... you can ignore what he says
biggrin.gif


PS. Raja stated that the gaming Vega tweaks specific to RX Vega not on the Vega frontier cards as well.
Thought i was going crazy lol
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top