AMD's Vega-based Cards to Reportedly Launch in May 2017 - Leak

[Blueberries]

If AMD can release a dual-Vega card with HBM and a closed loop cooler around $600 or so it would blow nVidia's gaming market out of the water until 2018.

It's not likely though

(I'm assuming Vega 10 will beat Pascal in Perf/watt but not in throughput)

 

[efikkan]

(I'm assuming Vega 10 will beat Pascal in Perf/watt but not in throughput)

Vega will not beat Pascal in efficiency. Pascal is currently 80% more efficient than Polaris, there is no way AMD can improve like that over night.
And if Vega were more efficient, it would scale past Pascal.

 

[ensabrenoir]

If AMD can release a dual-Vega card with HBM and a closed loop cooler around $600 or so it would blow nVidia's gaming market out of the water until 2018.

It's not likely though

(I'm assuming Vega 10 will beat Pascal in Perf/watt but not in throughput)

...dual gpu cards have been quite cringe worthy (295X2 & TitianZ) and haven't benefited either team in quite a while..... unless you bought them for under $400.......even then you'd still be that odd fellow (kidding both gpus have a cult like following and a serious fan club)

 

[TheGuruStud]

Vega will not beat Pascal in efficiency. Pascal is currently 80% more efficient than Polaris, there is no way AMD can improve like that over night.
And if Vega were more efficient, it would scale past Pascal.

The new silicon alone is going to match/beat Pascal. It just gets worse with dx12/vulkan.

And 80%? Wut?

 

[efikkan]

The new silicon alone is going to match/beat Pascal. It just gets worse with dx12/vulkan.

How? Please explain yourself. If true it would be the greatest achievement ever, and a much larger progress than anything in over a decade, in fact it would require about the same improvement as ~HD 4000 vs Polaris in a single jump without any node shrinks, dream on…

 

[dalekdukesboy]

The new silicon alone is going to match/beat Pascal. It just gets worse with dx12/vulkan.

And 80%? Wut?

Admittedly this. I was kinda reading everyone saying 1080 if lucky probably not even 1080ti or titan certainly and I'm kind of thinking to myself yeah, we don't know for sure but seems awfully cynical for a GPU that AMD seems pretty confident is worth leaking info about and has promising physical specs plus the die shrink alone if all put together properly could be super efficient and fast.

 

[Captain_Tom]

I've not seen the details of Vega 11 yet, but I assume it will be GDDR 5(X).

HBM has so far been a bad move for AMD, and it's not going to help Vega 10 for gaming either. GP102 doesn't need it, and it's still going to beat Vega 10. HBM or better will be needed eventually, but let's see if even Volta needs it for gaming. AMD should have spent their resources on the GPU rather than memory bandwidth they don't need.

Nope it's HBM2. RX 470 - RX Fury will ALL be HBM2. HBM isn't expensive anymore, Nvidia is just milking their customers.


All you need to do is remember back to the HD 4870 that cost HALF as much as the 280. It had that "Expensive and new" GDDR5.

 

[TheGuruStud]

How? Please explain yourself. If true it would be the greatest achievement ever, and a much larger progress than anything in over a decade, in fact it would require about the same improvement as ~HD 4000 vs Polaris in a single jump without any node shrinks, dream on…

Take the 480 down to the reports of 95w with the refresh. Even if a little generous that matches its competition. Vega will have IPC gains (potentially good gains) and using same silicon.

 

[Captain_Tom]

Admittedly this. I was kinda reading everyone saying 1080 if lucky probably not even 1080ti or titan certainly and I'm kind of thinking to myself yeah, we don't know for sure but seems awfully cynical for a GPU that AMD seems pretty confident is worth leaking info about and has promising physical specs plus the die shrink alone if all put together properly could be super efficient and fast.

Have you looked at the leaked specs? It's easily 2-3x stronger than the 480, and that puts it firmly in Titan XP territory.


Now I don't have a crystal ball, but that's how strong the Vega 10 card SHOULD be. If it isn't, it will be a complete failure unless it uses like 100w.

 

[dalekdukesboy]

Have you looked at the leaked specs? It's easily 2-3x stronger than the 480, and that puts it firmly in Titan XP territory.


Now I don't have a crystal ball, but that's how strong the Vega 10 card SHOULD be. If it isn't, it will be a complete failure unless it uses like 100w.

Um...yes I looked at the specs, and I think you misread me entirely. I was saying I'm not sure why people are putting the card down and seems to me like it should be treated with great optimism and looking at those specs and die shrink gains in efficiency YEAH I'm saying I think it could be quite the card. Again I don't know for sure no crystal ball here either but I agree with you I think this will be serious competition for whatever Nvidia has out there right now if it is done right.

 

[Captain_Tom]

How? Please explain yourself. If true it would be the greatest achievement ever, and a much larger progress than anything in over a decade, in fact it would require about the same improvement as ~HD 4000 vs Polaris in a single jump without any node shrinks, dream on…

lmao are you high?!


The 1080 for reference is only 15% stronger than the Fury X. So you think AMD will have trouble making a card more than 15% stronger than their 1.5 year old flagship?


Dude stop drinking Nvidia's marketing koolaid...

 

[Captain_Tom]

Um...yes I looked at the specs, and I think you misread me entirely. I was saying I'm not sure why people are putting the card down and seems to me like it should be treated with great optimism and looking at those specs and die shrink gains in efficiency YEAH I'm saying I think it could be quite the card. Again I don't know for sure no crystal ball here either but I agree with you I think this will be serious competition for whatever Nvidia has out there right now if it is done right.

Buddy I was backing you up lol. I am on your side, sorry if it sounded like I was attacking you.

As for why people are so skeptical? I really am not sure, but my guess is Nvidia fanboys have just completely bought into Nvidia's marketing.


I mean at best Nvidia has a 33% efficiency advantage over AMD (Titan X vs 480). Just switching to HBM alone would make them equally efficient, and we already know that 14nm has become incredibly more efficient with recent maturity.

To compete with the Titan X AMD would have just had to scale up the 480 to a twice as big die. However they have done that, and then they overhauled the entire architecture. The RX Fury (Or 590?) should be a monster.

 

[efikkan]

Nope it's HBM2. RX 470 - RX Fury will ALL be HBM2. HBM isn't expensive anymore, Nvidia is just milking their customers.

So why will Vega 10 (the consumer version) only have 8 GB?
We all know the supply of HBM2 is low.

Take the 480 down to the reports of 95w with the refresh. Even if a little generous that matches its competition. Vega will have IPC gains (potentially good gains) and using same silicon.

There are mythical reports of super binnings of Polaris, but that's not a fair comparison.
This is a real world comparison:

(Note: Picture is cut)

 

[TheGuruStud]

So why will Vega 10 (the consumer version) only have 8 GB?
We all know the supply of HBM2 is low.


There are mythical reports of super binnings of Polaris, but that's not a fair comparison.
This is a real world comparison:
View attachment 83094
(Note: Picture is cut)

Now, you're just blatantly trolling and need banned.

 

[dalekdukesboy]

Buddy I was backing you up lol. I am on your side, sorry if it sounded like I was attacking you.

As for why people are so skeptical? I really am not sure, but my guess is Nvidia fanboys have just completely bought into Nvidia's marketing.


I mean at best Nvidia has a 33% efficiency advantage over AMD (Titan X vs 480). Just switching to HBM alone would make them equally efficient, and we already know that 14nm has become incredibly more efficient with recent maturity.

To compete with the Titan X AMD would have just had to scale up the 480 to a twice as big die. However they have done that, and then they overhauled the entire architecture. The RX Fury (Or 590?) should be a monster.

HA, that's funny, I almost said Hey buddy etc in response to you and said I'm on your side etc. I ended up rephrasing it but I saw your post and I thought it was mine because I was thinking it still and forgot which way I had typed it. Anyway I don't care about being "attacked" but thanks, I just misread your tone and admittedly was confused it sounded like you agreed but yet didn't, so yeah no worries. I like arguing my point anyway so no worries on my feelings lol. Yeah anyway I agree, I saw specs and the doom test and various other leaks and fact they are confident enough to leak it months in advance is a good sign by itself, and all the evidence points to a very strong high end GPU that finally will compete with Nvidia's high end offerings.

Also that last slide showing efficiency or lack thereof is well known about the rx480 I admit it's a disappointment but this is NOT rx480 no? If it were, we wouldn't even have this discussion of even the hint of potential AMD could compete, no one would argue for it and no one would bother arguing against it for it would be obvious to all it was like an AMD CPU nowadays and just ok for price point only not performance.

 

[efikkan]

Now, you're just blatantly trolling and need banned.

I'm getting tired of this old lie.
I've checked >60 RX 480 reviews, both the initial and the more recent ones, and they all show power usage of ~150W +/-10%, in fact Techpowerup's measurements is slightly below average. There are no significant difference between the first batches and the more recent ones, and there are no significant difference between reference and non-reference boards, just the usual variance you'll get with every card. What matters is what you'll get when you buy a card, not what one out of every thousand cards claims to perform. It's disingenuous to claim that buyers will get a "95W" card, when we all know they'll get a ~150W card.

 

[rruff]

Now, you're just blatantly trolling and need banned.

Do you have any legit info on 95W RX 480s? I'm having no luck with google.

 

[TheGuruStud]

I'm getting tired of this old lie.
I've checked >60 RX 480 reviews, both the initial and the more recent ones, and they all show power usage of ~150W +/-10%, in fact Techpowerup's measurements is slightly below average. There are no significant difference between the first batches and the more recent ones, and there are no significant difference between reference and non-reference boards, just the usual variance you'll get with every card. What matters is what you'll get when you buy a card, not what one out of every thousand cards claims to perform. It's disingenuous to claim that buyers will get a "95W" card, when we all know they'll get a ~150W card.

No, now you're BSing and changing the subject after being called out. A 480 does not use 80% more power than a 1060 (trying to slip in 1070/1080 as a replacement). You're not being disingenuous or ignorant. You're lying intentionally. STFU.

 

[kruk]

Vega will not beat Pascal in efficiency. Pascal is currently 80% more efficient than Polaris, there is no way AMD can improve like that over night.
And if Vega were more efficient, it would scale past Pascal.

Nice try:
https://www.techpowerup.com/reviews/Zotac/GeForce_GTX_1080_Amp_Extreme/30.html
Compare RX 470 with 1060 3GB and RX 480 with 1060 6 GB at 1080p.

 

[Captain_Tom]

So why will Vega 10 (the consumer version) only have 8 GB?
We all know the supply of HBM2 is low.

Why is anyone complaining about 8GB? That is definitely enough.

Furthermore let's see if AMD's new memory tech indeed cuts usage in half. If 8GB's of AMD acts as good as 16GB of Nvidia, Nvidia is the one who will need to step their game up.

 

[rruff]

Compare RX 470 with 1060 3GB and RX 480 with 1060 6 GB at 1080p.

The 1060 is 48% better perf/W than the 480. The 470 doesn't compete with the 1060, but it is better than the other AMD cards.

The 1050 Ti is 62% better than the 470, though. https://www.techpowerup.com/reviews/Zotac/GeForce_GTX_1060_Mini_3_GB/30.html

 

[Captain_Tom]

Nice try:
https://www.techpowerup.com/reviews/Zotac/GeForce_GTX_1080_Amp_Extreme/30.html
Compare RX 470 with 1060 3GB and RX 480 with 1060 6 GB at 1080p.

Bingo! If you remove cherry picking Nvidia, really isn't that far ahead.


It's also hilarious considering how GF's 14nm was clearly being rushed into usage before it was ready. Just look at how the "Power Hungry" Fury is only 15% behind the 1060. Pathetic Nvidia can barely stay ahead of AMD's 28nm

 

[kruk]

The 1060 is 48% better perf/W than the 480. The 470 doesn't compete with the 1060, but it is better than the other AMD cards.

The 1050 Ti is 62% better than the 470, though. https://www.techpowerup.com/reviews/Zotac/GeForce_GTX_1060_Mini_3_GB/30.html

If 470 doesn't compete with 1060, then 1050 Ti also doesn't compete with 470. Period!
Still waiting for the 80% difference ...

 

[rruff]

If 470 doesn't compete with 1060, then 1050 Ti also doesn't compete with 470. Period!
Still waiting for the 80% difference ...

Didn't say they did compete. The 470 is AMD's most efficient card, and the 1050 Ti, 1070, and 1080 all beat it by >60%. I think 60% is a lot.

And I'd love to see some verified testing of new Polaris cards that shows they really have 50% greater efficiency than when first introduced. The only thing I can find are rumors from 3 months ago.

 

[kruk]

Didn't say they did compete. The 470 is AMD's most efficient card, and the 1050 Ti, 1070, and 1080 all beat it by >60%. I think 60% is a lot.

And I'd love to see some verified testing of new Polaris cards that shows they really have 50% greater efficiency than when first introduced. The only thing I can find are rumors from 3 months ago.

But with the 1060 3GB that is closest with the performance it's sub 20% difference. What's your point?