Monday, October 29th 2018

AMD "Vega 20" GPU Not Before Late Q1-2019

AMD "Vega 20" is a new GPU based on existing "Vega" graphics architecture, which will be fabbed on the 7 nanometer silicon fabrication process, and bolstered with up to 32 GB of HBM2 memory across a 4096-bit memory interface that's double the bus-width of "Vega 10". AMD CEO Lisa Su already exhibited a mock-up of this chip at Computex 2018, with an word that alongside its "Zen 2" based EPYC enterprise processors, "Vega 20" will be the first 7 nm GPU. AMD could still make good on that word, only don't expect to find one under your tree this Holiday.

According to GamersNexus, the first "Vega 20" products won't launch before the turn of the year, and even in 2019, one can expect product launches till the end of Q1 (before April). GamersNexus cites reliable sources hinting at the later-than-expected arrival of "Vega 20" as part of refuting alleged "Final Fantasy XV" benchmarks of purported "Vega 20" engineering samples doing rounds on the web. Lisa Su stressed the importance of data-center GPUs in AMD's Q3-2018 earnings call, which could hint at the possibility of AMD allocating its first "Vega 20" yields to high-margin enterprise brands such as Radeon Pro and Radeon Instinct.
Source: GamersNexus (YouTube)
Add your own comment

39 Comments on AMD "Vega 20" GPU Not Before Late Q1-2019

#1
Shamalamadingdong
So you're running with the story that Vega 20 which has only been announced as a Radeon Instinct card will be late but won't be late in actuality?

Click bait 2/8. Fabricated story 5/7.
Posted on Reply
#2
PerfectWave
yeah all knows was not for the gaming market......
Posted on Reply
#3
nguyen
yah who knows maybe Vega 50 is supposed to be a gaming card that coming out in 2022 to compete with RTX 4050 ti
Posted on Reply
#4
Vayra86
nguyen, post: 3931265, member: 168196"
yah who knows maybe Vega 50 is supposed to be a gaming card that coming out in 2022 to compete with RTX 4050 ti
That is actually more likely than us seeing a Vega 20 compete anywhere in the gaming space.
Posted on Reply
#5
Jism
I think AMD is redesigning that chip again in order to compete with Nvidia's arsenal of new chips.
Posted on Reply
#6
ppn
They better be redesigning it, but nah. They are just facing delays and yelds.
Posted on Reply
#7
Assimilator
Jism, post: 3931296, member: 91255"
I think AMD is redesigning that chip again in order to compete with Nvidia's arsenal of new chips.
Redesign it how? Vega is GCN pushed as far as GCN can go, there is literally no way to redesign it without throwing the whole thing out and starting afresh, aka Navi. That's why AMD is in such a tough spot in their GPU segment, because they held off on designing GCN's successor for two generations too long, which is how we got Fury and Vega; and now they don't have anything post-Vega, so all they can do is respin Vega on a more efficient node in the hope that will keep the product in the headlines and help with the terrible heat and power consumption.
Posted on Reply
#8
kings
So, 7nm GPUs are "around the corner" for how many months now? 6-7 months?

What a long corner indeed!
Posted on Reply
#9
Vya Domus
Assimilator, post: 3931325, member: 7058"
Vega is GCN pushed as far as GCN can go, there is literally no way to redesign it without throwing the whole thing out and starting afresh
I love these comments.

How do you know that ? You worked at AMD designing GCN and know every nook and cranny of it in order to tell us that it can't be redesigned?
Posted on Reply
#10
efikkan
This may be unconfirmed rumors from friends of Steve, but doesn't sound too unlikely. At best we're hoping for something close to a paper launch in end of Q4 this year. The shipping volumes would be negligible anyway.

Jism, post: 3931296, member: 91255"
I think AMD is redesigning that chip again in order to compete with Nvidia's arsenal of new chips.
Apparently you don't know how long it takes to design a chip.
Just from tapeout of the final design to market, it usually takes one year or more. Any extra tweaks or bugfixes adds 4-5 months. End to end it takes 3-5 years to design a new chip, depending on how much is redesigned.

kings, post: 3931339, member: 180022"
So, 7nm GPUs are "around the corner" for how many months now? 6-7 months?

What a long corner indeed!
The transition to 7 nm is going to be slow. Even when the first chips arrive next year, shipping volumes will be limited throughout most of 2019, if not even longer.

AMD's next consumer lineup (Navi) is scheduled for mid to late 2019, if things go according to plan.
Posted on Reply
#11
Divide Overflow
Vega continues to be focused at the data centers. It might be keeping them afloat, but it's not endearing them to performance gamers.
Posted on Reply
#12
bug
which could hint at the possibility of AMD allocating its first "Vega 20" yields to high-margin enterprise brands such as Radeon Pro and Radeon Instinct
I must have missed stuff, but has there been one piece of news saying Vega20 was intended for gaming?
And yes, I know thr difference between Vega and Vega Pro is mostly in the driver, but has AMD ever stated officially that they plan to do the same with Vega 20?
Posted on Reply
#13
ppn
VEGA 7nm is in the same 2017-2020 roadmap. standing between VEGA 14nm, NAVI 7nm and NEXT GEN 7+. So yes it is for gaming. This is happening.
Posted on Reply
#14
bug
ppn, post: 3931457, member: 159444"
VEGA 7nm is in the same 2017-2020 roadmap. standing between VEGA 14nm, NAVI 7nm and NEXT GEN 7+. So yes it is for gaming. This is happening.
Volta was on the same roadmap with Pascal, yet that didn't make it a gaming card ;)
I know there's a good chance Vega 20 will make it to the consumer space, I was just asking if this was ever confirmed by AMD or it's still our wishful thinking and/or educated guess.
Posted on Reply
#15
Shamalamadingdong
ppn, post: 3931457, member: 159444"
VEGA 7nm is in the same 2017-2020 roadmap. standing between VEGA 14nm, NAVI 7nm and NEXT GEN 7+. So yes it is for gaming. This is happening.
That doesn't mean anything. It's an architecture roadmap. Architecture != Gaming products.
Posted on Reply
#16
Valantar
bug, post: 3931437, member: 157434"
I must have missed stuff, but has there been one piece of news saying Vega20 was intended for gaming?
And yes, I know thr difference between Vega and Vega Pro is mostly in the driver, but has AMD ever stated officially that they plan to do the same with Vega 20?
Nope, you're spot on. Vega 20 has been all-but-confirmed to be non-consumer only. That this article lists it as having 32GB of HBM2 should make that plenty clear for anyone, even if it's of course possible to cut that to 16 or even 8, the arch makes no sense for a consumer card. It'll be a HPC/workstation/datacenter beast, but nothing for gaming. As expected. A good strategic choice by AMD in their current position, even if it leaves gamers wanting more than mainstream performance in the clammy hands of Nvidia.

Don't see how this is news, really.
Posted on Reply
#17
lynx29
nguyen, post: 3931265, member: 168196"
yah who knows maybe Vega 50 is supposed to be a gaming card that coming out in 2022 to compete with RTX 4050 ti
lol gg
Posted on Reply
#18
prtskg
Valantar, post: 3931498, member: 171585"
Nope, you're spot on. Vega 20 has been all-but-confirmed to be non-consumer only. That this article lists it as having 32GB of HBM2 should make that plenty clear for anyone, even if it's of course possible to cut that to 16 or even 8, the arch makes no sense for a consumer card. It'll be a HPC/workstation/datacenter beast, but nothing for gaming. As expected. A good strategic choice by AMD in their current position, even if it leaves gamers wanting more than mainstream performance in the clammy hands of Nvidia.

Don't see how this is news, really.
Exactly, anyone with a sane mind can see that it's not a consumer product.
Posted on Reply
#19
WikiFM
Assimilator, post: 3931325, member: 7058"
Redesign it how? Vega is GCN pushed as far as GCN can go, there is literally no way to redesign it without throwing the whole thing out and starting afresh, aka Navi. That's why AMD is in such a tough spot in their GPU segment, because they held off on designing GCN's successor for two generations too long, which is how we got Fury and Vega; and now they don't have anything post-Vega, so all they can do is respin Vega on a more efficient node in the hope that will keep the product in the headlines and help with the terrible heat and power consumption.
As far as I know Navi is still GCN (6th gen), Arcturus will be the one based on a new arquitecture, that till 2020 at best.
Posted on Reply
#20
prtskg
WikiFM, post: 3931549, member: 176404"
As far as I know Navi is still GCN (6th gen), Arcturus will be the one based on a new arquitecture, that till 2020 at best.
Arcturus is a chip name, not an architecture name. AMD is reverting to it's old way of naming chips by way of code name rather than name with some numbers like Vega 10 or Polaris 10. Source is Bridgeman from AMD on Phoronix.
Posted on Reply
#21
efikkan
ppn, post: 3931457, member: 159444"
VEGA 7nm is in the same 2017-2020 roadmap. standing between VEGA 14nm, NAVI 7nm and NEXT GEN 7+. So yes it is for gaming. This is happening.
AMD have stated on several occasions that Vega 20 is a professional product, even as recent as this summer. Plans are always subject to change, but Vega 20 is not intended as a consumer product.
Posted on Reply
#22
WikiFM
prtskg, post: 3931563, member: 163162"
Arcturus is a chip name, not an architecture name. AMD is reverting to it's old way of naming chips by way of code name rather than name with some numbers like Vega 10 or Polaris 10. Source is Bridgeman from AMD on Phoronix.
Well you are right, still Arcturus will be a chip based in a new nameless architecture different than Navi (that's still GCN).

I think Vega 20 is too expensive to produce to be competitive in the gaming market, that is why it won't launch in any consumer products till yields improve enough to be mass produced at a good price point.
Posted on Reply
#23
Prima.Vera
Looks more and more like AMD is repeating the 3DFx history all over again...
Posted on Reply
#24
rvalencia
Vya Domus, post: 3931350, member: 169281"
I love these comments.

How do you know that ? You worked at AMD designing GCN and know every nook and cranny of it in order to tell us that it can't be redesigned?
AMD is stuck with four CU lanes with 64 ROPS setup since R9-290X design. Vega 64 has four CU lanes and 64 ROPS with L2 cache link and 1536 Mhz clock speed range.

AMD moved from two CU lanes with 32 ROPS 7970 in December 2011 to four CU lanes with 64 ROPS R9-290X in year 2013. From 2013 to 2018, there was no increase in CU lane and ROPS count, while NVIDIA GPUs scaled towards 88 to 96 ROPS.

Raja "TFLOPS" Koduri joined AMD in 2013.
Posted on Reply
#25
Vayra86
ppn, post: 3931457, member: 159444"
VEGA 7nm is in the same 2017-2020 roadmap. standing between VEGA 14nm, NAVI 7nm and NEXT GEN 7+. So yes it is for gaming. This is happening.
As usual you are very crafty at pulling facts and specs out of your rear end.

Source or it didn't happen, most of what you're saying only happens in your head.

rvalencia, post: 3931952, member: 99935"
AMD is stuck with four CU lanes with 64 ROPS setup since R9-290X design. Vega 64 has four CU lanes and 64 ROPS with L2 cache link and 1536 Mhz clock speed range.

AMD moved from two CU lanes with 32 ROPS 7970 in December 2011 to four CU lanes with 64 ROPS R9-290X in year 2013. From 2013 to 2018, there was no increase in CU lane and ROPS count, while NVIDIA GPUs scaled towards 88 to 96 ROPS.

Raja "TFLOPS" Koduri joined AMD in 2013.
Nice find and linking of those facts. It will be interesting putting those next to what Raja's going to produce for Intel.

GCN was already up for a radical redesign in 2013 in every way and Hawaii was already scaled up too far, its a diminishing returns fiesta. At that point it was rapidly losing ground in perf/watt, ran into heat problems, and lacked optimization and focus compared to the competition. Its clear AMD has done everything in the book to stretch the 'old' GCN out for several more years with minimal effort and investment (and make it viable for emerging markets, too, quite a feat if you think of it, its the only way their HBM focus makes sense). Does that mean it is end of life? Not sure you can say that, it still has potency, but it fails to extract that proper for specific use cases (such as gaming). The way I see it, GCN can only survive when it gets more specialized, narrowed down and split up into branches for specific markets.
Posted on Reply
Add your own comment