• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Vega Rumors

CSI_PC said:
Here is a more reliable video showing mining for air and watercooled Vega with current existing drivers; air at 36Mh/s (hitting clock rates similar to what Tom's review saw in games when initially cold with around 1400Mhz) and 41Mh/s watercooled (this was hitting 1600MHz then 1520Mhz but considering this is from cold not suprising but with tweaking can it be maintained after 30mins).


Like he says, drivers should improve this a bit more, but this test is with cold GPUs so there is a balance between the two factors.
Cheers
Click to expand...

Maybe that is why they are putting the liquid cooled ones behind the bundle paywall to discourage miners from snapping them all up.
 
CSI_PC said:
Here is a more reliable video showing mining for air and watercooled Vega with current existing drivers; air at 36Mh/s (hitting clock rates similar to what Tom's review saw in games when initially cold with around 1400Mhz) and 41Mh/s watercooled (this was hitting 1600MHz then 1520Mhz but considering this is from cold not suprising but with tweaking can it be maintained after 30mins).


Like he says, drivers should improve this a bit more, but this test is with cold GPUs so there is a balance between the two factors.
Cheers
Click to expand...

That's the founder's edition:

Check out his other videos. It's not RX Vega.

https://www.youtube.com/channel/UC9W2WQsQLM9WmfqsMcht9Pw
 
Some accurate indicators from air cooled Vega FE and its envelope of clocks/TDP/thermals.
When cold clocks were 1440MHz with 293W, while when hot after awhile game benchmarking clocks were 1269MHz with 266W.
Those clocks for air this puts it around the 1080FE performance level (importantly the custom AIB 1080s probably a bit in front) in more recent DX12 games while DX11 will be a challenge to compete at times with the 1080FE even with newer drivers unless they can improve sustained clock rates closer to boost limits - so to me this makes the water cooled model possibly the only route to go for a fair few here if looking only at Vega in context of hardcore gaming and assuming it is binned giving best performance and efficiency envelope out of the range (seemed FuryX was when looking at accurate power measurements compared to rest of the Fiji range).


aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9TL0YvNjkzMDg3L29yaWdpbmFsLzAwLVBvd2VyLURyYXctT3ZlcnZpZXcucG5n



aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9TL00vNjkzMDk0L29yaWdpbmFsLzA4LVRlbXBlcmF0dXJlLXZzLUNsb2NrLVJhdGUtdnMtUGVyZm9ybWFuY2UucG5n


http://www.tomshardware.com/reviews/amd-radeon-vega-frontier-edition-16gb,5128-11.html

Just focused on this because drivers will not improve TDP nor clocks greatly, and this can give us a stronger indicator where each of the range is going to end up.
This does ignore memory clocks, but looks like there is very limited OCing capability with the HBM2 implementation.
Cheers
 
Last edited:
CSI_PC said:
I linked his very latest video, the link you give just shows the list and no others apart from the latest one I linked.
Cheers
Click to expand...
I linked ALL his videos because he has:
part i, with Vega FE
part ii with Vega FE
and part iii (which you linked) - which assumably is still Vega FE. It's not light gray in the video, like it is on his intro shot -- and I didn't watch the whole thing through - but he does not mention it is RX Vega.
 
Archaea said:
I linked ALL his videos because he has:
part i, with Vega FE
part ii with Vega FE
and part iii (which you linked) - which assumably is still Vega FE. It's not light gray in the video, like it is on his intro shot -- and I didn't watch the whole thing through - but he does not mention it is RX Vega.
Click to expand...
Yes but I edited to make my point clearer, this is the latest video and do you honestly expect the prosumer Vegas to perform worst than Vega consumer??
So this is with a bit of time since he did it 2 weeks ago.

Something would be wrong if the prosumer compute orientated GPU performs worst than the gaming Vega for mining....
Also like I pointed out, the clocks are hitting the TDP/TBP limits as measured by Tom's hardware and especially the air model, so again not sure how gaming Vega card is going to do better than the Vega FE unless they break their spec.

So alarm bells should be raised when a new rumour comes along saying gaming Vega is hitting 75+Mh/s, because a) memory clocking is limited due to HBM2 b) the core clocks are running into their TDP spec limits and so limiting gains there.
Both of which would be applicable to all Vega models in context of mining, also gaming but that was not the focus.
Cheers
 
Last edited:
*puts on tin foil hat*

Maybe this is part of AMD's plan, delay 2 weeks to build up stock, generate some hype over a fake 70-100MH/s.... cash out!

I'd be pretty impressed personally.
 
There is 1 thing we know for sure, RX Vega at the same clocks as a Vega FE will use less power. Why? It will only have 8GB of HBM2, compared to 16GB on the FE.

How much power saving that is, is anyones guess. Maybe 20-30w difference? anyone chime in on this?
 
Brackle said:
There is 1 thing we know for sure, RX Vega at the same clocks as a Vega FE will use less power. Why? It will only have 8GB of HBM2, compared to 16GB on the FE.

How much power saving that is, is anyones guess. Maybe 20-30w difference? anyone chime in on this?
Click to expand...
That's probably accurate if they don't clock the ram any higher.

It still doesn't account for tweaks to AVFS and the power saving features. Those are still getting regular updates with the linux drivers and likely not fully enabled with FE. These patches are likely landing concurrently with Windows driver development as AMD was sharing most of the work between platforms.

Also some recent Linux benches. Looks like AMD has managed a good 25-50% increase on 580/Fury in last month or so with drivers. Almost ahead of Windows gaming performance. Keep in mind a lot of these titles were ported into OpenGL and not originally optimized for Linux or using a wrapper.
http://www.phoronix.com/scan.php?page=article&item=radeonsi-beats-wingl&num=3

Marketing guys seem to think there is a lot of potential for improvement as well.
 
Brackle said:
There is 1 thing we know for sure, RX Vega at the same clocks as a Vega FE will use less power. Why? It will only have 8GB of HBM2, compared to 16GB on the FE.

How much power saving that is, is anyones guess. Maybe 20-30w difference? anyone chime in on this?
Click to expand...

Well that is why they said 285 watt TDP for the aircooled version of Vega RX. Its a bit lower than the air cooled FE which was at 300 watts.
 
Man RAJA hypes things so much. Only thing I don't like about him. Vega 20 year vision! WTF! lol. He doesn't know what he says because people are going to shred his words apart. He made it seem like he is going to shrink vega for next 20 years, lol. SMH.
 
razor1 said:
They have much work to do with their Linux drivers, always been that way, don't see that changing anytime soon.
Click to expand...
Economies of size. AMD struggled with their DX11 drivers on windows, so it would be crazy to think they have the manpower to invest in strong linux support. Right now I am sure their driver teams are focused on two things Vulkan and DX12 support and supporting Apple as I am sure that is a cash cow.
 
Presbytier said:
Economies of size. AMD struggled with their DX11 drivers on windows, so it would be crazy to think they have the manpower to invest in strong linux support. Right now I am sure their driver teams are focused on two things Vulkan and DX12 support and supporting Apple as I am sure that is a cash cow.
Click to expand...


completely agree!
 
Anarchist4000 said:
Marketing guys seem to think there is a lot of potential for improvement as well.
Click to expand...


This is so embarrassing for them! The morons are basically admitting they can't optimize for their hardware fast enough and expect their users to wait years for incremental improvements..LOL! They are basically marketing the "Fine Wine" bullshit made up by the sheep over at r/amd.
 
5150Joker said:
This is so embarrassing for them! The morons are basically admitting they can't optimize for their hardware fast enough and expect their users to wait years for incremental improvements..LOL! They are basically marketing the "Fine Wine" bullshit made up by the sheep over at r/amd.
Click to expand...


its like buying a stock and hoping it will go up in price (stead of taking the sure bet by researching a stock) to find out in 4 years ya get 10%, when you should be making 10% per year ;).
 
razor1 said:
its like buying a stock and hoping it will go up in price (stead of taking the sure bet by researching a stock) to find out in 4 years ya get 10%, when you should be making 10% per year ;).
Click to expand...
But it was a very efficient 10%
 
Maddness said:
Hopefully like the title says. It could be FUD. Gamers don't need this.
Click to expand...
I don't think AMD designed this with just gamers in mind nor do I think they care if lots of miners buy them. Techreport has a good write on the architecture and features and Vega seems to have mining in mind.
 
Presbytier said:
Economies of size. AMD struggled with their DX11 drivers on windows, so it would be crazy to think they have the manpower to invest in strong linux support. Right now I am sure their driver teams are focused on two things Vulkan and DX12 support and supporting Apple as I am sure that is a cash cow.
Click to expand...
They've had 8 guys doing the Linux work according to Raja in an interview somewhere. Regardless they've made a ton of progress and it wouldn't be surprising to see Linux overtake Windows in the next year. Big need for Linux gaming is Vulkan which is finally arriving in most major engines and DX12 can be ported rather easily.

As for operating systems, there's more than just Windows and Linux that may be relevant.

5150Joker said:
This is so embarrassing for them! The morons are basically admitting they can't optimize for their hardware fast enough and expect their users to wait years for incremental improvements..LOL! They are basically marketing the "Fine Wine" bullshit made up by the sheep over at r/amd.
Click to expand...
Devs can't really push out some of the features yet. Not until the new shader model hits without leaning on intriniscs. Already have Vega56 performing around GTX1080 with liquid cooled Vega64(more hardware and higher clocks) a good deal above that. Doubt that's even bothering with packed math and other optimizations.
 
Presbytier said:
I don't think AMD designed this with just gamers in mind nor do I think they care if lots of miners buy them. Techreport has a good write on the architecture and features and Vega seems to have mining in mind.
Click to expand...

I agree and i don't particularly like it.
 
Maddness said:
Hopefully like the title says. It could be FUD. Gamers don't need this.
Click to expand...
Maybe, but if AMD can ramp up production and sell mining cards for $1k each they'll make a killing.

Presbytier said:
I don't think AMD designed this with just gamers in mind nor do I think they care if lots of miners buy them. Techreport has a good write on the architecture and features and Vega seems to have mining in mind.
Click to expand...
Mining isn't the likely explanation for the high hash rate. The cards would have been designed for AMDs SME (memory encryption) as a security feature on the platform. Hence really efficient acceleration of SHA and AES algorithms.
 
Anarchist4000 said:
They've had 8 guys doing the Linux work according to Raja in an interview somewhere. Regardless they've made a ton of progress and it wouldn't be surprising to see Linux overtake Windows in the next year. Big need for Linux gaming is Vulkan which is finally arriving in most major engines and DX12 can be ported rather easily.

As for operating systems, there's more than just Windows and Linux that may be relevant.
Click to expand...

8 people on one driver project is a good size team ;) Its very hard to scale up on apps like drivers with more team members there is only so much work that can be done because how how the driver team would need split up based on the work split up.

Devs can't really push out some of the features yet. Not until the new shader model hits without leaning on intriniscs. Already have Vega56 performing around GTX1080 with liquid cooled Vega64(more hardware and higher clocks) a good deal above that. Doubt that's even bothering with packed math and other optimizations.
Click to expand...

This will have to go into my sig lol! come on man, AMD themselves show that isn't going to happen........
 
Anarchist4000 said:
They've had 8 guys doing the Linux work according to Raja in an interview somewhere. Regardless they've made a ton of progress and it wouldn't be surprising to see Linux overtake Windows in the next year. Big need for Linux gaming is Vulkan which is finally arriving in most major engines and DX12 can be ported rather easily.

As for operating systems, there's more than just Windows and Linux that may be relevant.
Click to expand...
It would take a divine intervention for Linux to over take Windows in the next year.
 
Anarchist4000 said:
Maybe, but if AMD can ramp up production and sell mining cards for $1k each they'll make a killing.


Mining isn't the likely explanation for the high hash rate. The cards would have been designed for AMDs SME (memory encryption) as a security feature on the platform. Hence really efficient acceleration of SHA and AES algorithms.
Click to expand...


At 1k a pop the cards have to get 90 mhs to be as profitable as current cards the power consumption can also be absorbed too with that hash rate.

PS alt coins don't use SHA or AES algorithms either (well they do, but they aren't profitable enough to really mine, and ASIC miners blow GPU's out of the water for hash rates on those too). SHA is used in Bitcoin though and its difficulty rating is too high for GPU miners current rate of return on SHA ASIC's is 6 months per unit (for the best one!), and they are doing multiple terahashs a sec.
 
Last edited:
Chimpee said:
From an 1070 killer to a 1080 killer, boy I wish I have his optimism......
Click to expand...


If it was that capable AMD would be all over it making their teams work day in and night out to get it done and pay triple overtime!
 
Anarchist4000 said:
Already have Vega56 performing around GTX1080 with liquid cooled Vega64(more hardware and higher clocks) a good deal above that. Doubt that's even bothering with packed math and other optimizations.
Click to expand...
LOL, that's a funny one, mark my word though, Vega 56 is not going to be substantially faster than a 1070, those numbers you read in that rumor are fake or min fps (they are too low for a 1070), AMD would have been up and about about Vega 56 if they were true. And we've already seen Vega 64 numbers, barely faster than a 1080FE (which boosts lower than aftermarkets 1080s). You either have Vega 56 so close to Vega 64 that 64 becomes irrelevant, or these numbers are false. So nobody believes these Vega 56 results, except those grasping for straws. And rest assured, I am gonna remind you of that very soon.
 
Last edited:
Chimpee said:
From an 1070 killer to a 1080 killer, boy I wish I have his optimism......
Click to expand...
What optimism? That's precisely what the leaks showed and the expected performance.

What's odd is the shear level of denial some of the posters have around here. Half a dozen guys running around hysterically trying to pretend they didn't screw up big time again with everyone thinking their idiots. I'm still laughing at the fact AMD played them so easily.
 
You must log in or register to reply here.
Back
Top