AMD Vega 10, Vega 11, Vega 12 and Vega 20 Rumors and Discussion

http://www.gamersnexus.net/news/1387-nvidia-drivers-337-vs-mantle-aggressive

There is also the issue that it seems like many of the new hardware / architecture features don't appear to be enabled / working properly, as tests have shown the tiled rasterization doesn't appear to be working and memory bandwidth is much worse than Fury X while it should be about the same w/o the features, and much faster with them working.

Anyway, its only a few weeks away now, and AMD might even give us a sneak peak during their tour: https://radeon.com/en-us/rx-community-meetup/
That's mostly SLI, which is even more blatantly dependant on a working driver profile.

This part of the presentation looks like it is the most realistic representation at least NV-internally:
http://media.gamersnexus.net/images/media/2014/nvidia/nv-dx6.jpg

When taking a look at it - even under aspects of supposedly better CPU-utilization - we were less than thrilled:
http://www.pcgameshardware.de/Nvidi...-sGeforce-Treiber-33750-gegen-Mantle-1116527/
Except for the starswarm techdemo, which showed tremendous gains, there was little overall improvement to be had for single-GPU.
 
On another note: Is there any indication of how high HBM gen2 temperatures are allowed to be without the chips throttling and/or suffering from severe degradation of signal quality?
 
I don't think theres enough black magic in the world to make the car 40% faster via drivers not just in a couple of months but in its entire life span.
Well, plot thickens. AMD is going to "discuss the latest and greatest from Radeon Software […] bring you the updates and plethora of features from this major Radeon Software release". There are two NDA dates, some info will be released a few days before Siggraph, the rest during Siggraph.
 
With Vega not performing better that Fiji clock for clock in games, but better in "pro" benchmark, I guess the driver thing is a *small* possibility after all. Something doesn't fit. The driver suck or the hardware is broken. I hope it's the first option.
 
With Vega not performing better that Fiji clock for clock in games, but better in "pro" benchmark, I guess the driver thing is a *small* possibility after all. Something doesn't fit. The driver suck or the hardware is broken. I hope it's the first option.
I guess we all do.
 
With Vega not performing better that Fiji clock for clock in games, but better in "pro" benchmark, I guess the driver thing is a *small* possibility after all. Something doesn't fit. The driver suck or the hardware is broken. I hope it's the first option.
No, best chance (for vega&amd) is if these pro-cards are from 1st chip revision which was faulty somehow, and game-cards will be later, fixed one (Barcelona TLB-like)
 
No, best chance (for vega&amd) is if these pro-cards are from 1st chip revision which was faulty somehow, and game-cards will be later, fixed one (Barcelona TLB-like)

I doubt they can do a hardware fix / respin in a few weeks. Or, the Vega in the FE edition are pretty old (like January or something), and the "fixed" one in RX Vega are just out of the fab. I don't believe in that right now...
 
Anybody noticed the absence of any form of comments from @sebbbi on the matter? Usually he has at least some preliminary thoughts. I mean he has the Vega FE now, he is the better equipped guy to make at least some educated guesses? Are his hands tied with an NDA? Even Rys had had to make some comments and he is working for RTG!
 
Maybe he's not even using it ? He seems busy working on "his" engine if you follow him on twitter. I doubt he has time for testing a new card right now if he doesn't have a benefit to do it ?
 
Some rumblings of Vulkan 1.1 at SIGGRAPH which should be a similar toolset to DX12's SM6.0. Some developers likely have different drivers and documentation under NDA.
 
Anybody noticed the absence of any form of comments from @sebbbi on the matter? Usually he has at least some preliminary thoughts. I mean he has the Vega FE now, he is the better equipped guy to make at least some educated guesses? Are his hands tied with an NDA? Even Rys had had to make some comments and he is working for RTG!
IIRC he said sometime ago he's waiting for the RX
 
Well, plot thickens. AMD is going to "discuss the latest and greatest from Radeon Software […] bring you the updates and plethora of features from this major Radeon Software release". There are two NDA dates, some info will be released a few days before Siggraph, the rest during Siggraph.
Where did you learn that?
 
Hey, I'm "bootgras" from reddit :). I've been checking this thread out for a while and finally decided to make an account, haha.

As for as Vega FE's temps/OC/etc:
  • The 1440mhz average clock speed is basically 100% due to the power limit. +35% is necessary to maintain 1600mhz+.
  • The FE's fan is more than capable of cooling the card, but it's limited to some nearly silent speeds by default (2000rpm). Kinda strange because under 3500rpm it's not any louder than a typical blower card. This is the secondary reason that it won't stay at 1600mhz, but the power limit is the main thing.
  • Overclocking/undervolting is simple...seems like the people that reviewed this card can't figure out wattman. I wonder how they deal with changing BIOS settings if they find wattman so difficult? :\
  • HBM2 temps are the real limiter for FE when stock... Once they hit 90C, the card throttles. And they hit 90C pretty easily. Have to keep GPU temps below 80 to keep it from running away... I hope there is some voltage control or power management that can deal with this in the future. Doesn't look like GN bothered to check how it's doing with their modded card.
 
  • HBM2 temps are the real limiter for FE when stock... Once they hit 90C, the card throttles. And they hit 90C pretty easily. Have to keep GPU temps below 80 to keep it from running away... I hope there is some voltage control or power management that can deal with this in the future. Doesn't look like GN bothered to check how it's doing with their modded card.

Interesting thanks! Wonder how much this will matter with custom coolers since coolers like the Fury Nitro were amazing at keeping it cool while almost silent still.
 
Hey, I'm "bootgras" from reddit :). I've been checking this thread out for a while and finally decided to make an account, haha.

As for as Vega FE's temps/OC/etc:
  • The 1440mhz average clock speed is basically 100% due to the power limit. +35% is necessary to maintain 1600mhz+.
  • The FE's fan is more than capable of cooling the card, but it's limited to some nearly silent speeds by default (2000rpm). Kinda strange because under 3500rpm it's not any louder than a typical blower card. This is the secondary reason that it won't stay at 1600mhz, but the power limit is the main thing.
  • Overclocking/undervolting is simple...seems like the people that reviewed this card can't figure out wattman. I wonder how they deal with changing BIOS settings if they find wattman so difficult? :\
  • HBM2 temps are the real limiter for FE when stock... Once they hit 90C, the card throttles. And they hit 90C pretty easily. Have to keep GPU temps below 80 to keep it from running away... I hope there is some voltage control or power management that can deal with this in the future. Doesn't look like GN bothered to check how it's doing with their modded card.
Well, 4GB HBM2 stacks should be easier to cool than 8GB stacks, so there's that.
 
On another note: Is there any indication of how high HBM gen2 temperatures are allowed to be without the chips throttling and/or suffering from severe degradation of signal quality?

AMD's proposals for more tightly integrated HBM have a stated desire to stay at 85C or below. Above 85C, the DRAM must begin to increase its refresh rate to compensate for the capacitors in the arrays leaking charge more readily. Some of AMD's TOP-PIM research proposed a curve of doubling the refresh rate for every 10C above the threshold.
It would start to impact the availability of the DRAM rather significantly, and harm perf/W.

The more recent spec for HBM that includes what is labelled HBM2 leaves a fair amount up to the device vendor, although a temperature-compensating refresh mode is present. I think HBM originally couldn't readily measure past 90-95C, but it seems like the latest thermal readouts can max out at ~125C, although that sounds like it's in the realm of shutting down due to the GPU's not being that tolerant and because HBM2 has an emergency shut-down sensor that would take things down at some point.

Anybody noticed the absence of any form of comments from @sebbbi on the matter? Usually he has at least some preliminary thoughts. I mean he has the Vega FE now, he is the better equipped guy to make at least some educated guesses? Are his hands tied with an NDA? Even Rys had had to make some comments and he is working for RTG!
Rys gave him the card, and as the legend goes it was done over a beer. He said further disclosures would be up to Rys.
I'm pretty sure agreements over a beer rank somewhere in the hierarchy of oath-swearing methods used in myth.
Sure, it could be broken, but "an hour of wolves, and shattered shields", etc.

  • HBM2 temps are the real limiter for FE when stock... Once they hit 90C, the card throttles. And they hit 90C pretty easily. Have to keep GPU temps below 80 to keep it from running away... I hope there is some voltage control or power management that can deal with this in the future. Doesn't look like GN bothered to check how it's doing with their modded card.
The HBM2 is trying to dissipate heat when it is right next to a 220W+ source, and its main thermal contact is the copper base that is being heated by said GPU perhaps a mm away.
It would be natural even without it's own power dissipation that it would rise to the ambient temperature, which the GPU would under load be raising to 80-90. There may be a limit to what it can do, since the big controlling factor is how heavily the GPU is using memory.

We could possibly consider this an additional secondary weakness of AMD's choice to use HBM, particularly with GPUs that are so much less efficient.
8-hi stacks may make things warmer for some of the layers versus what a 4-hi could do, but I've wondered before if there were something that could be done to give the stacks a contact with less direct influence from the GPU without making the hop to a closed loop cooler.
 
Maybe he's not even using it ? He seems busy working on "his" engine if you follow him on twitter. I doubt he has time for testing a new card right now if he doesn't have a benefit to do it ?
Anybody noticed the absence of any form of comments from @sebbbi on the matter? Usually he has at least some preliminary thoughts. I mean he has the Vega FE now, he is the better equipped guy to make at least some educated guesses? Are his hands tied with an NDA? Even Rys had had to make some comments and he is working for RTG!
The Vega conversation become too heated for me. I know that people are very excited about this card. I assumed there would be lots of reviews of Vega FE at launch, but everybody started to ask me to do a review instead. I even got requests from press to come and benchmark to our office. There's no conspiracy or anything. I am just letting AMD and the press to do their work, while focusing on my own.

I am currently finalizing our tech and I have my hands full of tasks. I skipped E3 and SIGGRAPH this year to have more time to get things done. Our launch is getting closer and shipping to 5 platforms (base & upgraded consoles + PC) with a 3 people team is pretty hard.

Vega is working well and I use it daily. Async compute is also working fine. I had plans to test some fp16 code on it, but I found out that Unreal Engine doesn't yet support fp16 on PC. So I am waiting for Epic to add fp16 support before I continue with fp16 work on PC. Ran my buffer benchmark on Vega (https://github.com/sebbbi/perftest), and noticed no surprises. It behaves like any other GCN based GPU (same peak perf/clock on various L1$ buffer load operations). Our game is mostly ray-traced, so it has no very little overdraw. Tiled rasterizers shouldn't help much at all, so I haven't spend any time confirming gains on that regard. I will test DX 12.1 features (especially volume tiled resources) after our game launches.

The card is very quiet. Even when playing our game at 4K without vsync. Running it at stock settings (like our CPUs and other components). In gamedev hardware stability is always more important than slight perf increase. I'd assume AMD has plenty of headroom to run the fan faster in consumer release if they want to.
 
AMD's proposals for more tightly integrated HBM have a stated desire to stay at 85C or below. Above 85C, the DRAM must begin to increase its refresh rate to compensate for the capacitors in the arrays leaking charge more readily. Some of AMD's TOP-PIM research proposed a curve of doubling the refresh rate for every 10C above the threshold.
It would start to impact the availability of the DRAM rather significantly, and harm perf/W.

The more recent spec for HBM that includes what is labelled HBM2 leaves a fair amount up to the device vendor, although a temperature-compensating refresh mode is present. I think HBM originally couldn't readily measure past 90-95C, but it seems like the latest thermal readouts can max out at ~125C, although that sounds like it's in the realm of shutting down due to the GPU's not being that tolerant and because HBM2 has an emergency shut-down sensor that would take things down at some point.
That's what I was concerned about. Thank you. So maybe maybe maybe, 4-hi-stacks will have more thermal leeway in some way or another.
 
Back
Top