AMD Vega Discussion Thread

[Th3pwn3r]

Right, it's not available yet for sale BUT reviewers have had their hands on them.

 

[P4-630]

Right, it's not available yet for sale BUT reviewers have had their hands on them.

That's why I thought you could buy them already...
Only saw the Gigabyte 64 selling in my country sofar.

 

[londiste]

1) Just by oc the HBM2 up to 950MHz gives 14,7% more!

not really. note that in addition to hbm oc, power limit was also lifted.
comparing that to stock+50% pwr, the difference is ~7%. not very good for 18% overclock on hbm (800-950mhz).

edit:
actually, their articles say 3.6-4.3% (no idea why it's different every time).

 

[TheoneandonlyMrK]

not really. note that in addition to hbm oc, power limit was also lifted.
comparing that to stock+50% pwr, the difference is ~7%. not very good for 18% overclock on hbm (800-950mhz).

edit:
actually, their articles say 3.6-4.3% (no idea why it's different every time).

800-950 my memory is at 945 stock ?? Dunno what gives here is it mine at higher than stock or ??

Ah its a 56 no problems

 

[turbogear]

I own Sapphire Radeon RX Vega 64 air cooled version since Thursday last week.
I had to pay 649€ for it. This is a bundled version. It comes with two games here in Germany (Prey and Sniper Elite 4).
As I did not own any of these two games before so it was ok for me.
In effect I paid only 309€ for Vega as I was able to sell my two year old Radeon Fury X for 340€ on eBay.

The 509€ version of Vega 64 was sold out within minutes after Mindfactoy put it online!
Here in Germany there are still some cards on stock at Mindfactory: https://www.mindfactory.de/Hardware/Grafikkarten+(VGA)/Radeon+RX+VEGA.html

I will change the blower fan cooler very soon. I have ordered "EK-FC Radeon Vega - Nickel" water block which should arrive in the next days.

Does anybody know if it would be posible to flash the bios of RX Vega liquid cooled version onto the aircooled version?
With proper water cooling the aircooled version in theory should reach the performance of Liquid cooled versions.
Of course this is not 100% guaranteed as Liquid cooled version from AMD has higher binning, but I would like to try it.

 

[TheoneandonlyMrK]

I own Sapphire Radeon RX Vega 64 air cooled version since Thursday last week.
I had to pay 649€ for it. This is a bundled version. It comes with two games here in Germany (Prey and Sniper Elite 4).
As I did not own any of these two games before so it was ok for me.
In effect I paid only 309€ for Vega as I was able to sell my two year old Radeon Fury X for 340€ on eBay.

The 509€ version of Vega 64 was sold out within minutes after Mindfactoy put it online!
Here in Germany there are still some cards on stock at Mindfactory: https://www.mindfactory.de/Hardware/Grafikkarten+(VGA)/Radeon+RX+VEGA.html

I will change the blower fan cooler very soon. I have ordered "EK-FC Radeon Vega - Nickel" water block which should arrive in the next days.

Does anybody know if it would be posible to flash the bios of RX Vega liquid cooled version onto the aircooled version?
With proper water cooling the aircooled version in theory should reach the performance of Liquid cooled versions.
Of course this is not 100% guaranteed as Liquid cooled version from AMD has higher binning, but I would like to try it.

That binning thing is not garunteed in some cases it would make sense to put a lower bin under an oem waterblock , for example low yeilds might dictate the requirement of the better bin to make air cooled clock to a point wherein the water cooled lower binned leaky chip is fine under water anyway , im not saying this is whats gone on because I don't know but I can think of a way to find out.

 

[Vya Domus]

Does anybody know if it would be posible to flash the bios of RX Vega liquid cooled version onto the aircooled version?

According to @buildzoid , flashing a different BIOS simply does not work , at least from what I understand.

 

[TheoneandonlyMrK]

According to @buildzoid , flashing a different BIOS simply does not work , at least from what I understand.

Could be similar to Polaris in that the drivers was selective about working or not if it noted bios changes , for Polaris this happens with modded Rx 480 bios but not if it's a 580 bios , and with some drivers not all, a pixel patcher patch got round that so there is hope

 

[turbogear]

According to @buildzoid , flashing a different BIOS simply does not work , at least from what I understand.

I also heard this too concerning the moded bios flashing on Vega, but I am not too sure if this is also the case if one tries the official AMD liquid Vega 64 bios on Vega 64 air cooled.
I thought the Vega 64 air cooled and liquid cooled are basically the same hardware but with different cooling solution and different bios.

I would like to try it in anycase, if I get somewhere the bios for the liquid cooled version.
TPU database does not contain any Vega bioses so far. The cards are too new.

 

[the54thvoid]

I also heard this too concerning the moded bios flashing on Vega, but I am not too sure if this is also the case if one tries the official AMD liquid Vega 64 bios on Vega 64 air cooled.
I thought the Vega 64 air cooled and liquid cooled are basically the same hardware but with different cooling solution and different bios.

I would like to try it in anycase, if I get somewhere the bios for the liquid cooled version.
TPU database does not contain any Vega bioses so far.

I would strongly suggest you do not flash the bios on your Vega card. It's been suggested that voltages and binning has been done and that alone is enough to think there could be issues. Wait for an overclocking fix, assuming their is one soon. You don't bios flash before trying standard overclocking first. If that route is broken, there could be a very good reason for it.
If your card is improperly binned for aggressive clocks on higher voltages, it will end in tears.

 

[cucker tarlson]

Apart from the horrendous thermals and loads of heat it dumps into the room, Vega is bugged with stuttering in games and the clock bug. Not a card I would like to run in my system even if it was 50% off the original fake msrp. This is clearly a compute card, not king of the budget gaming GPUs, AMD doesn't give a rat's ass about gamers.

 

[turbogear]

I would strongly suggest you do not flash the bios on your Vega card. It's been suggested that voltages and binning has been done and that alone is enough to think there could be issues. Wait for an overclocking fix, assuming their is one soon. You don't bios flash before trying standard overclocking first. If that route is broken, there could be a very good reason for it.
If your card is improperly binned for aggressive clocks on higher voltages, it will end in tears.

Thanks for the advice. I will then wait for an overclocking fix.
I don't want to destroy my new card with improper bios.

My new water block is coming soon and I will see how does the clock behaves ones the card is properly cooled.
I expect that my Vaga will stay at higher clocks for longer time with proper water cooling.
With the original cooler and default fan speed setting the card hits 85°C very quickly and Wattman sends out warnings all the time, if you leave Wattman open in background.
I manually dropped the Max temperature to 75°C and increased the fan speed. The average clock speed was higher but the GPU sounded like a Vacuam cleaner!
I usually game with closed headset on. It did not disturb me too much but the temperatures on the card remained average in 70°C range.

 

[Vya Domus]

I also heard this too concerning the moded bios flashing on Vega, but I am not too sure if this is also the case if one tries the official AMD liquid Vega 64 bios on Vega 64 air cooled.
I thought the Vega 64 air cooled and liquid cooled are basically the same hardware but with different cooling solution and different bios.

I would like to try it in anycase, if I get somewhere the bios for the liquid cooled version.
TPU database does not contain any Vega bioses so far. The cards are too new.

Could be similar to Polaris in that the drivers was selective about working or not if it noted bios changes , for Polaris this happens with modded Rx 480 bios but not if it's a 580 bios , and with some drivers not all, a pixel patcher patch got round that so there is hope

Yes , Polaris would validate the VBIOS through the driver. However Vega seems to have actual hardware locking/BIOS checking.

I expect that my Vaga will stay at higher clocks for longer time with proper water cooling.
With the original cooler and default fan speed setting the card hits 85°C very quickly and Wattman sends out warnings all the time, if you leave Wattman open in background.
I manually dropped the Max temperature to 75°C and increased the fan speed. The average clock speed was higher but the GPU sounded like a Vacuam cleaner!
I usually game with closed headset on. It did not disturb me too much but the temperatures on the card remained average in 70°C range.

I would honestly suggest running the card on the power saver mode , seems to reduce power consumption and temps drastically while losing little performance. You can also place a custom fan profile for the time being and make it even less noisy.

 

[HD64G]

not really. note that in addition to hbm oc, power limit was also lifted.
comparing that to stock+50% pwr, the difference is ~7%. not very good for 18% overclock on hbm (800-950mhz).

edit:
actually, their articles say 3.6-4.3% (no idea why it's different every time).

Correct! So, 5% for VRAM oc (might change to bigger % when on bigger resolutions) and 9% for power limit +50% which will be seen easily for custom cards on default or +20% at most due to better thermals. Asus Strix rpreviews strongly comfirm that. So, +15% easily without stretching its legs at all. This brings it close to stock 1080 than to 1070. Not bad value at all eh?

 

[EarthDog]

Indeed.. however, the other card can overclock as well.

 

[cucker tarlson]

1080 FE is 2% faster than V64 Reference, GTX 1080 gaming 10Gbps is 3% faster than Vega Strix, GTX 1080 Gaming 11Gbps is 6% faster than Vega Strix. According to TPU's review, actual 3D performance gained from overclocking the GTX 1080 Gaming 11Gbps is 8.1%. So we're now at +15% over vega strix for the oc'd 1080 gaming. OC vs OC we're getting the same performance, but the power draw difference probably gets even bigger with GTX 1080 OC consuming 250W according to TPU while computerbase measures Vega 64 with +50% power at a whooping 406W, and that comes only with increasing the power limit, no overclocking or adding voltage. The 100W difference between Gaming X + and Vega 64 now grows to 160W at least, possibly close to 200W when adding voltage and increasing clocks. Kind of big, knowing that 1080Ti draws only about 60-70W more than 1080 and offers monstrous performance gains over the mid-sized chip.

Nvidia's performance/efficiency numbers for the big chips with 384-bit bus actually fall on the same level as their medium and small chips with 192/256-bit,which is incredible. I think there's still a way for Pascal to get even bigger, I just don't know if there's a will on nvidia's side cause it might not be profitable. A 300W card with +15% performance over 1080Ti ? Hell, I'd buy that for the winter season

 

[mastrdrver]

So I guess primitive discard is not enabled in any public driver. Straight from the horses mouth:

Ryan Smith said so on B3D forums

Quick note on primitive shaders from my end: I had a chat with AMD PR a bit ago to clear up the earlier confusion. Primitive shaders are definitely, absolutely, 100% not enabled in any current public drivers.

The manual developer API is not ready, and the automatic feature to have the driver invoke them on its own is not enabled.

 

[Vya Domus]

So I guess primitive discard is not enabled in any public driver. Straight from the horses mouth:

Ryan Smith said so on B3D forums

If that's true then I wonder why was Buildzoid getting some weird glitches at high clocks that looked like the effects of primitive discarding.

I always had a suspicion this wasn't as easy to implement as AMD made it sound like.

 

[TheoneandonlyMrK]

So I guess primitive discard is not enabled in any public driver. Straight from the horses mouth:

Ryan Smith said so on B3D forums

If that's true then I wonder why was Buildzoid getting some weird glitches at high clocks that looked like the effects of primitive discarding.

I always had a suspicion this wasn't as easy to implement as AMD made it sound like.

especially as they describe it just working with no input from devs , so that Rpm and tile based rendering in an as yet undefined state with no end goal date for switch on.

to be fair mines running quite well now , and plays most games at 4k ultra settings including a few more that two or three polaris i had setup could not manage in dx12 for example but two never mind three polaris beat this Vega in flat out fps when use able, three beat my mates 1080ti(@timespy and gtaV) but then that is three.

and yes fx8350 can just about do that but i cant wait to have some free cash for a cpu mobo mem combo.

I think buildzoids glitches are typical memory on the edge glitches tbh.

 

[HD64G]

I was almost sure that the new and important features were disabled in Vega RX drivers. So, much more potential in next few driver updates than the usual 5-10%. Interesting times ahead then...

 

[notb]

So I guess primitive discard is not enabled in any public driver. Straight from the horses mouth:

Ryan Smith said so on B3D forums

I wonder if they'll manage to activate it before the next generation arrives. Unbelievable. This card is so late and it still looks like something released half a year too early. Or it's actually the next gen card during development, but they were forced to put something on sale this year.

I can't believe NVIDIA didn't make an answer for Vega now.
They could make a dual 1070 (which would be like what... 50-60% faster than Vega 64 at the moment?) and just market it as "Hey, are you really fine with 350W of power draw? Get this. Vega might get better over time, but will never match it."

I was almost sure that the new and important features were disabled in Vega RX drivers. So, much more potential in next few driver updates than the usual 5-10%. Interesting times ahead then...

You're getting an unfinished product and you're actually praising AMD. :-/

 

[turbogear]

I can't believe NVIDIA didn't make an answer for Vega now.
They could make a dual 1070 (which would be like what... 50-60% faster than Vega 64 at the moment?) and just market it as "Hey, are you really fine with 350W of power draw? Get this. Vega might get better over time, but will never match it."
:-/

They don't need to make a new card. They have 1080Ti which much faster than Vega 64.

Either AMD is not able to make this feature work until now or they have too less resources. I wonder how large is the driver team at AMD.

 

[notb]

They don't need to make a new card. They have 1080Ti which much faster than Vega 64.

Either they are not able to make this feature work until now or they have too less resources. I wonder how large is the driver team at AMD.

1080Ti is about where Vega 64 might arrive with all the great drivers and optimization that we've been promised.
Dual 1070 would be significantly faster. In fact a pushed 1080Ti would also be faster. It's just that both options would go beyond 300W TDP and until Vega arrived it wasn't really considered acceptable. But now it looks like there really is a tiny interest in Vega, so NVIDIA could release a low volume product to quickly take over this niche.

And BTW: I've just checked the workstation Vega (WX9100) and the TDP is just 250W - that's called meeting industry standards. With an "enthusiast" product they could go totally bonkers with power consumption... so they did.

 

[EarthDog]

I was almost sure that the new and important features were disabled in Vega RX drivers. So, much more potential in next few driver updates than the usual 5-10%. Interesting times ahead then...

such a familiar tune...

 

[turbogear]

1080Ti is about where Vega 64 might arrive with all the great drivers and optimization that we've been promised.
Dual 1070 would be significantly faster. In fact a pushed 1080Ti would also be faster. It's just that both options would go beyond 300W TDP and until Vega arrived it wasn't really considered acceptable. But now it looks like there really is a tiny interest in Vega, so NVIDIA could release a low volume product to quickly take over this niche.

I own a Vega 64 and happy with it so far but somehow I am not currently believing that it would reach 1080Ti performace, especially compared to the OC editions of 1080Ti.
I would be very happy if it does, but lets wait and see what later drivers will bring.