The Fury X Owners Club

[the54thvoid]

I've made custom Fury and Fury X BIOSs. They have a 1.3V stock voltage, 400W power limit and 350A current limit. For me the BIOSs score better in benchmarks and they clock a little better than the stock BIOSs.

Here's a download link for all of them.

I will not make BIOSs for the Nano cards because I do not trust the VRM on those to not blow up.

The BIOSs were all made by Hexediting the BIOSs from the TPU database and my own Fury X BIOS. They also have custom IDs so it's easy to tell when it's flashed.

I don't own Fury cards but it's work like yours that helps keep enthusiasts going. Well done

 

[Fouquin]

I will not make BIOSs for the Nano cards because I do not trust the VRM on those to not blow up.

They would not blow up, but it's a bit of a waste of time due to the severe power limiting. Increasing voltage on a Nano decreases boost clock rate. Unless you can fully remove the throttling (which is hardware and firmware related) than any custom VBIOS for a Nano would only be to increase the power limit and clockrate to that of the Fury X, which most can do without any problems.

Important note about the Nano and overclocking, raw clockrate does not seem to be the deciding factor on performance. I ran mine at 1120MHz and it performed worse overall than at 1080MHz. Again, this is down to the balance of power delivery and how heavily utilized the core can be. With all that said, increasing the HBM clock speed always gave a performance increase.

If you were to make a custom Nano VBIOS keep it light. Increase power limits to allow for 200W of draw, increase GPU clock to 1050MHz, and increase HBM to 550MHz or 570MHz, but leave voltage alone. I'm sure that most cards will whine like crazy under load, but that should give a very respectable performance increase that doesn't require any further tuning.

 

[buildzoid]

They would not blow up, but it's a bit of a waste of time due to the severe power limiting. Increasing voltage on a Nano decreases boost clock rate. Unless you can fully remove the throttling (which is hardware and firmware related) than any custom VBIOS for a Nano would only be to increase the power limit and clockrate to that of the Fury X, which most can do without any problems.

Important note about the Nano and overclocking, raw clockrate does not seem to be the deciding factor on performance. I ran mine at 1120MHz and it performed worse overall than at 1080MHz. Again, this is down to the balance of power delivery and how heavily utilized the core can be. With all that said, increasing the HBM clock speed always gave a performance increase.

If you were to make a custom Nano VBIOS keep it light. Increase power limits to allow for 200W of draw, increase GPU clock to 1050MHz, and increase HBM to 550MHz or 570MHz, but leave voltage alone. I'm sure that most cards will whine like crazy under load, but that should give a very respectable performance increase that doesn't require any further tuning.

The card already has a 200W power limit. There's a secondary power limit which is at 150W which applies to the 8 pin and then the current limit is 270A just like the Fury X. Unless a free Nano shows up for me to test with I am not making a high power BIOS for the Nano because it's exactly the same core as a Fury-X but fed by a VRM that is 2/3rds of the Fury-Xs and I don't trust it. Also because the card has a single 8 pin there isn't much room for more power through the cable. If the card had 2 6 pins I'd do a 200W+ BIOS or something but with a single 8pin you can run the risk of melting the cable.

 

[Fouquin]

The card already has a 200W power limit.

Tested stock I was reading 188W maximum draw in 3DMark. 200W (+/- 4W) was with the power limit increased to +20%. +50% power limit allowed just 226W peak. The moment I increased voltage the power draw decreased to 209W due to throttling. All these ratings are from the wall, and I was running clocks of 1077/600 for all but the stock setting.

So again, a simple VBIOS edit to set the clocks to 1050/550 would be exceptionally easy and completely stable. Also interesting, decreasing core voltage by 14mv while leaving the power limit increased by +40% actually gave better performance across the board, if only by 2%.

Keep in mind I am NOT telling you to do it. Only sharing information based on the testing and tweaking I did with a Nano.

 

[buildzoid]

So I've modding the crap out of my 3 cards to try takes some HWbot high scores. Turns out that adding capacitors to the HBM VRM is very very effective at improving HBM overclocking ranges for benchmarks. even my worst card which would insta crash above 550mhz HBM now does 600mhz stable enough to finish Firestrike. It does artifact like crazy I don't think this is any use for daily gaming. Might be better with raised HBM voltage but I can't be bothered taking the card apart for the 2nd time(the stock cooler has way too many screws) when it might already be fast enough for the scores I need. Speaking of scores it managed 21.1K in the GPU test in Firestrike with Tesselation off(HWbot rules).

Here's a pic of how the worst card I have looks ATM:

Other than that I've made another batch of custom BIOSs. Now with support for the Windforce Fury and the Nitro Fury. There's also a new Fury X BIOS included based of the latest Fury X BIOS that AMD released. It gets better core clocks than the old one. I managed to bench at 1190mhz core with the new BIOS without any extra voltage in Trixx which is 20mhz higher than the old one so it should be pretty good for all uses.

Link to the BIOS here: https://www.dropbox.com/s/v18py3rhbef8rmx/ProFijiBIOSs.zip?dl=0

 

[blacktruckryder]

Add me to the club! I picked up a Sapphire Fury X a few days ago. Loving the card so far! As soon as I find my maximum overclock I'll post up some benchmark scores.

One issue I've run into is, Afterburner won't let me clock the core any higher than 1140. I can set the clock speed higher but it doesn't change in any game I run, 1140Mhz is the maximum the core will go.

 

[GhostRyder]

So I've modding the crap out of my 3 cards to try takes some HWbot high scores. Turns out that adding capacitors to the HBM VRM is very very effective at improving HBM overclocking ranges for benchmarks. even my worst card which would insta crash above 550mhz HBM now does 600mhz stable enough to finish Firestrike. It does artifact like crazy I don't think this is any use for daily gaming. Might be better with raised HBM voltage but I can't be bothered taking the card apart for the 2nd time(the stock cooler has way too many screws) when it might already be fast enough for the scores I need. Speaking of scores it managed 21.1K in the GPU test in Firestrike with Tesselation off(HWbot rules).

Here's a pic of how the worst card I have looks ATM:

Other than that I've made another batch of custom BIOSs. Now with support for the Windforce Fury and the Nitro Fury. There's also a new Fury X BIOS included based of the latest Fury X BIOS that AMD released. It gets better core clocks than the old one. I managed to bench at 1190mhz core with the new BIOS without any extra voltage in Trixx which is 20mhz higher than the old one so it should be pretty good for all uses.

Link to the BIOS here: https://www.dropbox.com/s/v18py3rhbef8rmx/ProFijiBIOSs.zip?dl=0

With those mods and such while using the stock cooler how are temps?

 

[Ferrum Master]

Turns out that adding capacitors to the HBM VRM is very very effective at improving HBM

If you would have used some Sanyo OSCON not some crappy Wurth Electronic(china based, they don't produce anything actually)... maybe it would end up better.

 

[jboydgolfer]

View attachment 66820

What do you score with 8xAA?
Not specifically your card even, just any fury. At extreme tess,1080,8xaa,fullscreen,no amd contrlol panel aa,or tesselation disabled,just optimization,or use application setting option set?

Im curious

 

[buildzoid]

If you would have used some Sanyo OSCON not some crappy Wurth Electronic(china based, they don't produce anything actually)... maybe it would end up better.

These are 7mohm ESR, 2000uF polymers and as far as I know they're pretty damn good.

With those mods and such while using the stock cooler how are temps?

With the fans at 100% the cards don't break 50C under 3Dmark.

 

[Ferrum Master]

These are 7mohm ESR, 2000uF polymers and as far as I know they're pretty damn good.With the fans at 100% the cards don't break 50C under 3Dmark.

You need ripple current in this scenario if really the stock VRM is such a piece of crap and the output is unstable and noisy. But I don't trust any rebrand part makers... once you have this, another batch is from other place being crap. Nothing beats Oscon in SMPS crap filtering job IMHO.

 

[buildzoid]

You need ripple current in this scenario if really the stock VRM is such a piece of crap and the output is unstable and noisy. But I don't trust any rebrand part makers... once you have this, another batch is from other place being crap. Nothing beats Oscon in SMPS crap filtering job IMHO.

The ripple current on these is rated at 6.64A which is the highest I could find at this price point. All the Nichicon caps that were 5mohm ESR only did about 5A ripple current and came in smaller capacities.

The Vcore had 20mv of noise on it. I never checked the HBM VRM since I didn't think it would be very noisy since the HBM is so low power. I still slapped some caps on the HBM anyway since I bought 30 of them and was really surprised when 600mhz started to work.

 

[Ferrum Master]

Nichicon caps that were 5mohm ESR

ESR isn't the only thing... and each series of caps have their purpose of life... except good for nothing general purpose ones

Solder out some Sanyo and Panasonic ones from burned motherboard VRM for such experiments. The ripple is bigger because of having only one phase for the supply guess poorly filtered, because the card is smaller . Good catch btw.

 

[buildzoid]

ESR isn't the only thing... and each series of caps have their purpose of life... except good for nothing general purpose ones

Solder out some Sanyo and Panasonic ones from burned motherboard VRM for such experiments. The ripple is bigger because of having only one phase for the supply guess poorly filtered, because the card is smaller . Good catch btw.

Yeah that's what I figured. The HBM VRM has pads in place for a second phase and there is a lot of unused cap pads on it too.

I guess next time I'm ordering from digikey I'll buy a boatload of different caps and test which ones give me the best results. I have some old AMD GPUs with measly 3 phase Vcore VRMs that should see good gains from cap mods.

 

[blacktruckryder]

Here is my best run so far. 14009 http://www.3dmark.com/fs/8261609

 

[uuuaaaaaa]

Just got a R9 Fury Strix for 260 pounds, unlocked it to 3840SP's! ASIC quality is 60%...

 

[uuuaaaaaa]

Just unlocked to 4096 Shaders, I guess that I have a Fury X now!

https://www.techpowerup.com/gpuz/details/bkdfv

http://www.3dmark.com/fs/9806811

 

[ManofGod]

Just unlocked to 4096 Shaders, I guess that I have a Fury X now!

https://www.techpowerup.com/gpuz/details/bkdfv

http://www.3dmark.com/fs/9806811

Lucky you. I love my Sapphire Fury Nitro+ but, no unlocky for me.

 

[INSTG8R]

Well thanks to @gupsterg I am now a full "FuryX" on air!

 

[gupsterg]

Sweet , you done some stability testing?

Also I noticed your on Windows 10 from PM screenies, do you use CSM = Off on mobo bios to give you "pure UEFI" mode? if so for Fiji modded ROMs to work there is a custom UEFI/GOP module Lordkag made me on Fernando's Win-RAID forum.

 

[INSTG8R]

Sweet , you done some stability testing?

Also I noticed your on Windows 10 from PM screenies, do you use CSM = Off on mobo bios to give you "pure UEFI" mode? if so for Fiji modded ROMs to work there is a custom UEFI/GOP module Lordkag made me on Fernando's Win-RAID forum.

yeah I ran Firestrike and Time Spy on it so not really a stress test but the results spoke for themselves. I gonna thow some games at it.

I do have CSM "on" But I'm pretty sure it's mostly in legacy mode. I use Fast Boot setting in the BIOS anyway so I'm not bothered with not having Secure Boot. I'm on my desktop in 15 sec. anyway
Thanks for the all the help!

 

[gupsterg]

Yep, CSM=On = Legacy.

Sweet it passed FS & TS , you may have been super lucky with that card IMO . When I went for full 4096 SP unlock on my Tri-X I got desktop corruption so wasn't even able to load a 3D app .

 

[INSTG8R]

Yep, CSM=On = Legacy.

Sweet it passed FS & TS , you may have been super lucky with that card IMO . When I went for full 4096 SP unlock on my Tri-X I got desktop corruption so wasn't even able to load a 3D app .

Yeah I don't think many were sold when I bought mine so it was prolly a pretty "fresh" one that was just software disabled like the "4GB" 480s. I have chucked some DX:Mankind Divided at it and I will prolly fire up The Division, that usually makes things sweat.
Again Thanks for the help!

 

[gupsterg]

No worries glad help and super result .

If and when you OC, happy to help set/guide you on bios mod, it will make the card as if it was from factory the way you want .

Personally I prefered the Tri-X cooler for i) ease of card installation/removal vs Fury X with AIO ii) better GPU VRM temps . With a little fan profile mod in ROM I was able to gain pretty good temps for being just on air IMO.

 

[gupsterg]

Luv'in the Fiji even more .