AMD Vega 10, Vega 20, and Vega 11 GPUs Detailed

[Captain_Tom]

I was referring to the double-posting.

From the Forum Guidelines
"If you reply to multiple posts use the "multi quote" button, that way the forum is easier to read."

https://www.techpowerup.com/forums/threads/forum-guidelines.197329/

LOL I honestly didn't know about that. Thanks

If Vega 10 gets sold from 1st Q of 2017 they will have over 10 months until Volta gets on the market. And nobody knows which Volta will be 1st to come out after all. If not the biggest possible, Vega 10 could be too close to that also and beating 1080 for less money. How is that wrong for us? Only extreme gamers need 1080Ti and have the cash for it atm.

If the rumored specs are true it will absolutely crush the 1080 and trade blows or beat the Titan. The problem is timing, HBM2 selection, and pricing.

If AMD selects the dirt cheap 512 GB/s HBM2 then they better not price this more than $899, and I would hope they would choose 8GB to keep prices down. On the other hand if they select the 720 or 1000 GB/s HBM then they could put 16GB on it and charge $1000+.


Also timing is key. It seems like if they can launch an UBER card Before March they may have a full year before Nvidia can respond with Volta, and the worst they would have to put up with is some 1180 with 8GB of HBM in July. But if Vega launches in June it will give Nvidia supremacy of the high end for FAR too long - hopefully a 490 with GDDR5 (X?) will launch this fall either way.

 

[Vayra86]

LOL I honestly didn't know about that. Thanks

If the rumored specs are true it will absolutely crush the 1080 and trade blows or beat the Titan. The problem is timing, HBM2 selection, and pricing.

If AMD selects the dirt cheap 512 GB/s HBM2 then they better not price this more than $899, and I would hope they would choose 8GB to keep prices down. On the other hand if they select the 720 or 1000 GB/s HBM then they could put 16GB on it and charge $1000+.


Also timing is key. It seems like if they can launch an UBER card Before March they may have a full year before Nvidia can respond with Volta, and the worst they would have to put up with is some 1180 with 8GB of HBM in July. But if Vega launches in June it will give Nvidia supremacy of the high end for FAR too long - hopefully a 490 with GDDR5 (X?) will launch this fall either way.

Bonus points if you can see the irony here

 

[Captain_Tom]

Bonus points if you can see the irony here

Oh I can. I realized it the second I posted. Old habits die hard :/

 

[ensabrenoir]

If the rumored specs are true it will absolutely crush the 1080 and trade blows or beat the Titan. The problem is timing, HBM2 selection, and pricing.

If AMD selects the dirt cheap 512 GB/s HBM2 then they better not price this more than $899, and I would hope they would choose 8GB to keep prices down. On the other hand if they select the 720 or 1000 GB/s HBM then they could put 16GB on it and charge $1000+.


Also timing is key. It seems like if they can launch an UBER card Before March they may have a full year before Nvidia can respond with Volta, and the worst they would have to put up with is some 1180 with 8GB of HBM in July. But if Vega launches in June it will give Nvidia supremacy of the high end for FAR too long - hopefully a 490 with GDDR5 (X?) will launch this fall either way.

.....every Amd launch starts out like this......and then....

 

[Captain_Tom]

.....every Amd launch starts out like this......and then....

Nah the 290X overperformed.


The Fury X also had the correct specs and more-or-less performed as well as it should, it just needed more optimized drivers.

 

[$ReaPeR$]

And the 480, 470, and 460 are capturing marketshare right now. Polaris is allowing them to capture the biggest chunk of the market while they wait for API's and 14nm to mature. As AMD sees it there is no point in releasing Enthusiast cards if they won't be at their peak potential, and if developers won't fully utilize them.

well, yes, and i don't disagree with their move but they need something in the upper end of the market if they want to gain substantial market share because marketing works that way.

 

[Captain_Tom]

well, yes, and i don't disagree with their move but they need something in the upper end of the market if they want to gain substantial market share because marketing works that way.

I guess they are testing if it does.


They had plenty of marketshare with the weaker 4870, 5870 (This one did dominate for while), and 6970. Then when the 7970 and 290X were kicking ass they lost marketshare slowly but surely. Technically Radeon's best days were when they purely whent for price/perf and efficiency.

Having said that, the performance gulf between even a cut-down Vega and Polaris will be absolutely massive. I think they will be quite silly if they don't release at least an RX 490 this fall. They wouldn't even need a new die either, as long as 14nm is maturing they could just release P10 clocked at 1450MHz with GDDR5X and an 8-pin connector. That would give it a 20% boost over the 480 and allow it to take on the 1070 in at least DX12 games.

 

[Slizzo]

If Vega 10 gets sold from 1st Q of 2017 they will have over 10 months until Volta gets on the market. And nobody knows which Volta will be 1st to come out after all. If not the biggest possible, Vega 10 could be too close to that also and beating 1080 for less money. How is that wrong for us? Only extreme gamers need 1080Ti and have the cash for it atm.

Don't forget that nVidia is delivering Volta to the Gov. to fulfill a contract in early 2017. So it could come a lot sooner than people think if Vega turns out to be crazy. Though honestly I think it'll probably be around GTX1080/Titan X(P) performance.

 

[looncraz]

They wouldn't even need a new die either, as long as 14nm is maturing they could just release P10 clocked at 1450MHz with GDDR5X and an 8-pin connector. That would give it a 20% boost over the 480 and allow it to take on the 1070 in at least DX12 games.

Polaris 10 at just 1.3Ghz and GDDR5X would have a 20% improvement... the GPU is just that starving for bandwidth.

We can see that 1.6GHz GDDR5 (204GB/s) is the right amount for a 750MHz Polaris 10 GPU...

2.2GHz GDDR5 (280GB/s) would be ideal for 1Ghz P10 GPU...
2.8GHz GDDR5 (360GB/s) would be ideal for 1.25GHz P10 GPU
2.9GHz GDDR5 (370GB/s) would be ideal for 1.3GHz P10 GPU.

RX 480 has just 256GB/s of bandwidth... not ideal even at 1GHz GPU clocks.

Vega 10 will need about 75% more bandwidth to feed its CUs... so about 630GB/s would be "ideal."

EDIT:

Also, you can see the diminishing returns from GPU frequency from all of the tests I ran:

1250MHz and 1350MHz are much closer together than 1150MHz and 1250MHz...

 

[Captain_Tom]

Polaris 10 at just 1.3Ghz and GDDR5X would have a 20% improvement... the GPU is just that starving for bandwidth.

We can see that 1.6GHz GDDR5 (204GB/s) is the right amount for a 750MHz Polaris 10 GPU...

2.2GHz GDDR5 (280GB/s) would be ideal for 1Ghz P10 GPU...
2.8GHz GDDR5 (360GB/s) would be ideal for 1.25GHz P10 GPU
2.9GHz GDDR5 (370GB/s) would be ideal for 1.3GHz P10 GPU.

RX 480 has just 256GB/s of bandwidth... not ideal even at 1GHz GPU clocks.

Vega 10 will need about 75% more bandwidth to feed its CUs... so about 630GB/s would be "ideal."

EDIT:

Also, you can see the diminishing returns from GPU frequency from all of the tests I ran:

1250MHz and 1350MHz are much closer together than 1150MHz and 1250MHz...

That's actually really interesting, and similar to the gains I found overclocking 7970 memory (Stock was 1375, and even at 1850MHz the card clearly wanted more bandwidth).


I think AMD will have 2 Vega GPU's early next year:

-Cut-down 3584-SP with 768-GB/s HBM - $500 - $600

-Full-die 4096-SP with 1 TB/s HBM - $650 - $800

Keep in mind that nothing is ideal right now, so even if the chips are bandwidth starved AMD has to release something.

 

[looncraz]

Don't forget that nVidia is delivering Volta to the Gov. to fulfill a contract in early 2017. So it could come a lot sooner than people think if Vega turns out to be crazy. Though honestly I think it'll probably be around GTX1080/Titan X(P) performance.

Indeed, I think Volta will probably be focused on bringing compute improvements more than anything. This is allegedly a whole new architecture, though, which could mean anything from a gaming performance perspective.

 

[$ReaPeR$]

I guess they are testing if it does.


They had plenty of marketshare with the weaker 4870, 5870 (This one did dominate for while), and 6970. Then when the 7970 and 290X were kicking ass they lost marketshare slowly but surely. Technically Radeon's best days were when they purely whent for price/perf and efficiency.

Having said that, the performance gulf between even a cut-down Vega and Polaris will be absolutely massive. I think they will be quite silly if they don't release at least an RX 490 this fall. They wouldn't even need a new die either, as long as 14nm is maturing they could just release P10 clocked at 1450MHz with GDDR5X and an 8-pin connector. That would give it a 20% boost over the 480 and allow it to take on the 1070 in at least DX12 games.

for all of our sakes i hope you are right. the GPU market needs some competition and lower prices.

 

[Captain_Tom]

for all of our sakes i hope you are right. the GPU market needs some competition and lower prices.

Well we have great prices in the mid-range right now, period. But yeah Nvidia will continue to price gouge in the High-End as long as their only high-end competition is 1-year old Fury cards. Hopefully AMD launches the 490 this year.

The funny thing is AMD has also been able to price gouge -> in the Mid-Range. The 480 is beating the 1060 by quite a bit (It's still constantly selling out). As such the 480 is about $20-$40 more than it was supposed to be, and the 470 is being sold for the same price as lower-tier 480's since stock issues are such a problem. 4GB 460 prices are laughable, but again it is because it has zero competition in the space.

 

[$ReaPeR$]

Well we have great prices in the mid-range right now, period. But yeah Nvidia will continue to price gouge in the High-End as long as their only high-end competition is 1-year old Fury cards. Hopefully AMD launches the 490 this year.

The funny thing is AMD has also been able to price gouge -> in the Mid-Range. The 480 is beating the 1060 by quite a bit (It's still constantly selling out). As such the 480 is about $20-$40 more than it was supposed to be, and the 470 is being sold for the same price as lower-tier 480's since stock issues are such a problem. 4GB 460 prices are laughable, but again it is because it has zero competition in the space.

i dont know if this is AMD's fault or just a vendor issue since low numbers and high demand tend to have this effect on products.

 

[Captain_Tom]

i dont know if this is AMD's fault or just a vendor issue since low numbers and high demand tend to have this effect on products.

I really wish we could get some concrete numbers on GPU sales. All I have to go by is how often things are out of stock, and unverified leaks from vendors saying "We have 25x the 480's ready for launch compared to how many 1080's were ready for launch".

 

[$ReaPeR$]

I really wish we could get some concrete numbers on GPU sales. All I have to go by is how often things are out of stock, and unverified leaks from vendors saying "We have 25x the 480's ready for launch compared to how many 1080's were ready for launch".

well, yes, that's a tad difficult. but indeed it would clarify the situation.

 

[midnightoil]

Much of this 'news' is made up. At least the detail of it is.

NAVI is still expected for early 2018, a year after Vega.

Vega 20 (if it is to be called Vega) is a real part and is indeed 2019 ... but it won't be HBM2 32GB. Would make no sense whatsoever to have that much VRAM with that many shaders, unless 4K 144Hz / 8K is the norm by then, and the 7nm process take stock clocks to 2.5Ghz - 3Ghz. Neither of which is likely.

Vega 11 is in no way replacing Polaris 10. It will sit at or above Polaris 10, probably at or above Fury X performance. They'll be sold in the same segments that you'd expect them to inhabit today.

Vega 10 details are mostly right.

The thing that bothers me about Vega 10 & 11 is GF's copy (not very) exactly (in fact quite badly) Samsung 14nmFF LP+ process. If they're using it again for Vega (the high performance part), I expect power consumption and clocks may disappoint a little again.

We know Samsung are building something for AMD ... surely AMD should be using their new high performance 14nmFF process(es) for Vega 10, 11 and Zen?

 

[ValenOne]

What's with reducing the FP first from double to single, now half?
"with up to 24 TFLOP/s 16-bit (half-precision) floating point performance"
Obviously reducing the compute side increases gaming usability, as proven by Nvidia cards doing the same.

GP104's SM only has a single 16 bit FP unit with 128 32bit units which is different from GP100's SM design.

RX 480 has 5.8TFLOPS @ 1.266Ghz - GTX 1070 has 6.5TFLOPS at 1.683GHz. GTX 1060 has 4.4TFLOPS at 1.709GHz... and is about the same performance as RX 480 at 1.266GHz if both are pegged at the stated clocks (GTX 1060 usually runs at > 1.8GHz stock, RX 480 at ~1.2GHz stock, giving the GTX 1060 about a 10% lead in performance).

Vulkan/DX 12 are allowing AMD GPUs to fill in the gaps left by their scheduler windows... gaps which really shouldn't exist as much as they do in the first place. AMD needs a new ABI, scheduler, and driver in order to get rid of more of those gaps using DX11... but they could, conceivably, then see a ~25% boost in performance... without providing more hardware processing power...

Vega will need to come out with a low price, no doubt. This is one of the major reason I think AMD will use a 2048-bit HBM2 bus... two stacks are cheaper than four.. and even if the costs are not so much better, you still have a smaller die/interposer or both to help.

Volta will trash Vega 10, no doubt. The only way this wouldn't happen is if Vega is using the long-rumored new ABI... then we could see Vega 10 being
as much as twice as fast as the Fury X. I have serious doubts about this, though... the same team that created Polaris created Vega just six months later, so it's most certainly a direct a descendant of GCN 4... and there's some indication in my own circle that AMD has abandoned the new ABI altogether as the software is catching up and they anticipate DX11 performance to become irrelevant - as well as their utilization issues... all with no work on their part.

From https://developer.nvidia.com/dx12-dos-and-donts

On DX11 the driver does farm off asynchronous tasks to driver worker threads where possible.


NVIDIA's DX11 driver already has multiple threads and asynchronous tasks.

Much of this 'news' is made up. At least the detail of it is.


Vega 11 is in no way replacing Polaris 10. It will sit at or above Polaris 10, probably at or above Fury X performance. They'll be sold in the same segments that you'd expect them to inhabit today.

Vega 11 seems to be Scorpio's GPU build with it's worst working chip yield around 6 TFLOPS instead of Polaris 10's worst working chip yield of 4.9 TFLOPS for RX-470. Vega 11 effectively replaces Polaris 10.

For Scorpio, MS waited for AMD's medium size GPU chip to reach 6 TFLOPS with good yields.

Scorpio with Polaris 10 at 6 TFLOPS target would have terrible working chip yields.

 

[lorraine walsh]

DX11 is pretty good today with them mind you, they can only do so much with an architechture that REQUIRES synchronous loads, unlike Nvidia which is build to process specific loads altogether in one.

 

[Captain_Tom]

Much of this 'news' is made up. At least the detail of it is.


Vega 11 is in no way replacing Polaris 10. It will sit at or above Polaris 10, probably at or above Fury X performance. They'll be sold in the same segments that you'd expect them to inhabit today.

Why not?


A lot of reports stated that Samsung is working on making slower (768 GB/s) HBM2 chips that are far cheaper than HBM1. Having one unified architecture means they don't have to segment between GDDR and non-GDDR dies, which is incredibly helpful when it comes to utilizing early yields.

Furthermore it would shave off about 25 - 50w off of all of their models, and it would give them all a nice 20-50% performance boost from last years model.