Wednesday, August 30th 2017

AMD RX Vega 56 to Vega 64 BIOS Flash - No Unlocked Shaders, Improved Performance

A ChipHell forum user has done what probably others have already done in relative obscurity: trying (and succeeding) to flash a Vega 64 BIOS onto a Vega 56 graphics card. The result? Well, apparently the shaders won't unlock (at least not according to our very own GPU-Z), but interestingly, performance improves all the same. The lesser amount of shaders on the Vega 56 silicon (3585 Shaders / 224 TMUs / 64 ROPs compared to Vega 64's 4096 / 256 / 64 apparently doesn't hinder performance that much. It appears that the improved clockspeeds of Vega 56 after the BIOS flash do more than enough to offset performance loss from the lesser amount of compute resources available, bumping RX Vega's clock speeds of 1471 MHz core boost clock and 800 MHz HBM2 memory up to Vega 64's 1545 MHz core boost clock and 945 MHz HBM2 clock.

This means that Vega 56 can effectively become a Vega 64 in performance (at least where 3D Mark Fire Strike is concerned), which isn't unheard of in the relationship between AMD's top tier and second-best graphics cards. Now naturally, some Vega 56 samples may even be further overclocked than Vega 64's stock clocks, which means that there is the potential for Vega 56 to have even better performance than Vega 64. The BIOS swap should allow Vega 56 to access higher power states than its stock BIOS allows, which is one of the reasons it can unlock higher core and memory clocks than an overclocked, original BIOS Vega 56 would. However, the fact that a Vega 56 at Vega 64 clocks and a Vega 64 deliver around the same score in benchmarks definitely does raise questions on how well the extra computing resources of Vega 64 are being put to use.
Sources: ChipHell, via Videocardz
Add your own comment

60 Comments on AMD RX Vega 56 to Vega 64 BIOS Flash - No Unlocked Shaders, Improved Performance

#26
Raevenlord
News Editor
MrGenius

What is this? How am I supposed to know what that means?



Oh...I'm not supposed to know. I'm supposed to navigate to another site for a definition. Convenient.
Really? Honestly, I thought it was pretty clear myself. Some trouble decifering the "Vega 56" moniker under www.chiphell, but other than that...
Posted on Reply
#27
P4-630
Do people forget you can OC Vega 64 as well!?
If you OC Vega 64, I'm sure it will be faster than any OC'd/flashed Vega 56....
Posted on Reply
#28
Boosnie
This is good news.
There is much room to work on and 3rd party cards will be stellar.
Posted on Reply
#29
Hokum
I think we need to see how this scales with resolution. I wonder if the 64 is being bottle necked somewhere. But a 56 at 64 speeds being the same speed as a 64 does suggest inefficient usage.
Posted on Reply
#30
dyonoctis
P4-630Do people forget you can OC Vega 64 as well!?
If you OC Vega 64, I'm sure it will be faster than any OC'd/flashed Vega 56....
From what I'm understand, they just put the vega 56 at the same clock as the vega 64, and vega 56 still tied despiste having less core, that's what ppl are getting upset about, those 2 cards are awfully close, this result isn't logical, meaning that vega 64 isn't performing as well as it could. Even if you can oc vega 64 further, it doesn't change the fact that clock for clock vega 56 is just as good.

The only "good" thing is that not everyone will try this mod, fearing to brick their cards so vega 64 25% higher price over 56 won't upset people.
Posted on Reply
#31
Hokum
dyonoctisThe only "good" thing is that not everyone will try this mod, fearing to brick their cards so vega 64 25% higher price over 56 won't upset people.
Isn't there a BIOS switch on the Vega like my Fury? Just flash one and if you balls it up just flick the switch... There is little to lose...
Posted on Reply
#32
Durvelle27
NabarunEven if the 56 isn't immediately bricked, it definitely will lower the lifespan of the cards. The 64s are the creme-DE-la-creme of the chips - the 56s are not. They are not supposed to be taking it that hard. Push it and you get what you deserve.
Dude it will not decrease the life span any more than Overclocking would
Posted on Reply
#33
bug
This lends more credibility to the hypothesis Vega (and Fiji before it) is ROP limited. Something AMD has denied in the past (or at least claimed they "had no indication of").
A bit of information only useful to us, hardware freaks.
Posted on Reply
#34
SKD007
HD64GUndervolting and upping the power limit will lower temps and allow much higher clocks for long time me thinks. It has been tested elsewhere and works as well as with reference RX480, offering clearly better performance.
I already tried that my friend. No luck for me.
Posted on Reply
#35
TheinsanegamerN
LightofhonorYep, if AMD could dynamically reduce the cores running to increase clock speeds when needed, that would be awesome. Kinda like with CPU's having an all core boost and a 2-4 core boost.
Would that help though? in CPUs it helps because programs will often use either multiple slow or few fast cores.

Graphics scale with parallel cores much better then programs scale with CPU cores. I dont see how upping the clock rate and reducing core count would help in something that benefits from parallel processing.
Posted on Reply
#36
Nabarun
Durvelle27Dude it will not decrease the life span any more than Overclocking would
No. Changing the default firmware increases the voltage limits. That's the ONLY reason for anyone to try this. Feeding it more voltage than what AMD thinks is OK is what reduces the lifespan more than normal OC.
Posted on Reply
#37
Durvelle27
NabarunNo. Changing the default firmware increases the voltage limits. That's the ONLY reason for anyone to try this. Feeding it more voltage than what AMD thinks is OK is what reduces the lifespan more than normal OC.
Not enough of an concern when most users don't even keep GPUs more than 5 years

I myself upgrade yearly
Posted on Reply
#38
lightofhonor
TheinsanegamerNWould that help though? in CPUs it helps because programs will often use either multiple slow or few fast cores.

Graphics scale with parallel cores much better then programs scale with CPU cores. I dont see how upping the clock rate and reducing core count would help in something that benefits from parallel processing.
With Vega, a card held back by its own power and heat requirements, I think it would.
Posted on Reply
#39
Totally
NabarunNo. Changing the default firmware increases the voltage limits. That's the ONLY reason for anyone to try this. Feeding it more voltage than what AMD thinks is OK is what reduces the lifespan more than normal OC.
There isn't any difference between the dies their all the same the only thing separating the 56/64 is the gpu die itself, Thee 56 dies are most likely ones that initially didn't pass QA and have a defective CU likewise 64s are dies that didn't make the cut as an FE. The pcb and all other components are exactly the on both the Vega56 and Vega64. You are talking out your ass. It's clearly and painfully evident that the 56 is clocked lower because at the same clocks it performs nearly the same as the 64 and would cannibalize sales and make the 64 pointless.
Posted on Reply
#40
Nabarun
TotallyThere isn't any difference between the dies their all the same the only thing separating the 56/64 is the gpu die itself, Thee 56 dies are most likely ones that initially didn't pass QA and have a defective CU likewise 64s are dies that didn't make the cut as an FE. The pcb and all other components are exactly the on both the Vega56 and Vega64. You are talking out your ass. It's clearly and painfully evident that the 56 is clocked lower because at the same clocks it performs nearly the same as the 64 and would cannibalize sales and make the 64 pointless.
"talking out your ass" - Nice way to prove your intelligence. In the same sentence you admit the 56 is shittier than the 64 and yet claim it can handle more torture than it's supposed to. Just because you turned your 480 into 580 and haven't burnt it yet doesn't mean your ill-advice won't eat up others' cards. And BTW, a few MHz doesn't make a world of difference in gaming. Only difference it does make is significantly higher power usage, heat, noise and reduced lifespan of NOT ONLY the gpu but other components too- since the heat also affects other components in the case. IMHO, totally not worth it. But to each his own....
Posted on Reply
#41
Totally
Nabarun"talking out your ass" - Nice way to prove your intelligence. In the same sentence you admit the 56 is shittier than the 64 and yet claim it can handle more torture than it's supposed to. Just because you turned your 480 into 580 and haven't burnt it yet doesn't mean your ill-advice won't eat up others' cards. And BTW, a few MHz doesn't make a world of difference in gaming. Only difference it does make is significantly higher power usage, heat, noise and reduced lifespan of NOT ONLY the gpu but other components too- since the heat also affects other components in the case. IMHO, totally not worth it. But to each his own....
I'm sorry I was trying to communicate with you on your level because you are making some very wrong assumptions based on misconceptions. A defective CU does not necessarily have any effect on the voltage/current tolerances. "56 is shittier than the 64" no shit Sherlock, I was pointing out the obvious that seems to be oblivious to you. If you believe 64s are golden samples you are sorely mistaken as those are reserved OC editions or professional grade (i.e FE) where tolerances are tighter. Vega for gamers is the chaff, it pointless to bin it any further beyond a simple pass/fail. In regards to voltage, if the chip can't handle it at a certain point it won't stable, clearing that hurdle it's just a matter of cooling.
Posted on Reply
#42
Nabarun
0
TotallyI'm sorry I was trying to communicate with you on your level because you are making some very wrong assumptions based on misconceptions. A defective CU does not necessarily have any effect on the voltage/current tolerances. "56 is shittier than the 64" no shit Sherlock, I was pointing out the obvious that seems to be oblivious to you. If you believe 64s are golden samples you are sorely mistaken as those are reserved OC editions or professional grade (i.e FE) where tolerances are tighter. Vega for gamers is the chaff, it pointless to bin it any further beyond a simple pass/fail. In regards to voltage, if the chip can't handle it at a certain point it won't stable, clearing that hurdle it's just a matter of cooling.
Stability is not a guarantee of longevity. And temps don't kill - voltage does. May be not always immediately, but surely in time. It's up to the user to decide how much long he wants that to be. Cooling is not a solution for excessive voltages that sub-par chips are not built to handle. By your logic no chip should ever fry under Nitrogen. That's not the case, is it? Use some brains when you try to "communicate at the level of" a total stranger who you know nothing about.
Posted on Reply
#43
Vayra86
P4-630Do people forget you can OC Vega 64 as well!?
If you OC Vega 64, I'm sure it will be faster than any OC'd/flashed Vega 56....
Not when the 64 does not come with any headroom and the 56 does... Especially because of temps. And then there is a signficant price gap between these cards as well, in other words, the OC gain on the 64 does not pay off and the cheaper one does.
TotallyI'm sorry I was trying to communicate with you on your level because you are making some very wrong assumptions based on misconceptions. A defective CU does not necessarily have any effect on the voltage/current tolerances. "56 is shittier than the 64" no shit Sherlock, I was pointing out the obvious that seems to be oblivious to you. If you believe 64s are golden samples you are sorely mistaken as those are reserved OC editions or professional grade (i.e FE) where tolerances are tighter. Vega for gamers is the chaff, it pointless to bin it any further beyond a simple pass/fail. In regards to voltage, if the chip can't handle it at a certain point it won't stable, clearing that hurdle it's just a matter of cooling.
A cooler chip on high volts is still a chip on high volts that will degrade faster, VRM gets pushed harder too. Remember there is a board around a chip as well.
bugThis lends more credibility to the hypothesis Vega (and Fiji before it) is ROP limited. Something AMD has denied in the past (or at least claimed they "had no indication of").
A bit of information only useful to us, hardware freaks.
This has always been my take on it too.
TheinsanegamerNWould that help though? in CPUs it helps because programs will often use either multiple slow or few fast cores.

Graphics scale with parallel cores much better then programs scale with CPU cores. I dont see how upping the clock rate and reducing core count would help in something that benefits from parallel processing.
I do. Graphics go through a pipeline - if you saturate only some resources in the pipeline, while a portion of shaders is doing nothing but still running at high clock, you waste power. If you reduce the core clock globally, the entire pipeline slows down, even though you'd just want to touch the shaders. Alternatively, you could also build a more well balanced GPU to begin with (read: like Nvidia tightly balancing VRAM throughput with core performance and continuously shuffling around resources in the GPCs).
Posted on Reply
#44
EarthDog
Some interesting theories here... lol
Posted on Reply
#45
SKD007
I just flashed AIO sapphire bios to my AIr 64 and it sits at 70c all day stuck at 1300 to 1400.. but when I turn on the system when the card is cold, it hits 1750 and runs at 1640 something till it reaches 69 then bam it sits at 1300 to 1400 throttling for targeted 70c on the bios. So a custom liquid loop will help me go higher with air bios than the AIO bios I guess.
Posted on Reply
#46
ratirt
saikamaldossI just flashed AIO sapphire bios to my AIr 64 and it sits at 70c all day stuck at 1300 to 1400.. but when I turn on the system when the card is cold, it hits 1750 and runs at 1640 something till it reaches 69 then bam it sits at 1300 to 1400 throttling for targeted 70c on the bios. So a custom liquid loop will help me go higher with air bios than the AIO bios I guess.
Did you try lowering the voltage with same technique as with Vega 56?
Posted on Reply
#47
SKD007
ratirtDid you try lowering the voltage with same technique as with Vega 56?
Ya the problem is thermal limit. For LC bios, the thermal limit is 70 instead of 85 of AIr. It’s better to flash Air bios on LC and try Overclocking it. And may be I better wait for custom cooler bios from sapphire or Asus to try and flash that.. if I succeed then would buy ekwb water block and loop it to my h100i

As soon as card hits 69c then it drops from 1630 to 1300..
Posted on Reply
#48
ratirt
saikamaldossYa the problem is thermal limit. For LC bios, the thermal limit is 70 instead of 85 of AIr. It’s better to flash Air bios on LC and try Overclocking it. And may be I better wait for custom cooler bios from sapphire or Asus to try and flash that.. if I succeed then would buy ekwb water block and loop it to my h100i

As soon as card hits 69c then it drops from 1630 to 1300..
Did you try those 2 Bios options? I know there's been 2 of them. Maybe one is better than the other in terms of thermal's?
Posted on Reply
#49
SKD007
ratirtDid you try those 2 Bios options? I know there's been 2 of them. Maybe one is better than the other in terms of thermal's?
Which 2 options ?
Posted on Reply
#50
ratirt
saikamaldossWhich 2 options ?
I thought there are 2 Bios'es and as with the tests and benchmarks each bios used has had a different impact on the volts, power and heat. you have misunderstood me a bit. It's just 2 bios you can use and I was wondering if you tried both and saw same behavior with the frequencies.
Posted on Reply
Add your own comment
Dec 23rd, 2024 17:54 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts