Can't Overvolt 5870 Over 1.125v With RBE

[cadaveca]

Odd, 100%

As to multimeter readings, I have, yes(Assuming I'm using the right spot). It's the only reason I hate these backplates! volts area tad lower that what afterburner reads, and I get what seems to be a fair amount of droop, too. I'd love some cards with all the vgpu phases in place!

I'm running Crossfire, and that plus ULPS and Afterburner don't work well, so I have no other option but to flash bioses currently, if I want higher volts.

 

[misterpro]

That's really odd. And no kidding, not ideal!

My cards came with a 1.174v bios, and from the sounds of things, you got 1.162?

So those cards are the ones affected?

Misterpro here has 1.25v default?


WTH?

I wonder if Afterburner is actually raising the volts...
I've noticed that the actual clock profiling within the bios is very different on some cards...check out an XFX bios, then Sapphire, and ASUS, and you'll find they use different positions for 3D clocks...

Working bios is using "clock info 3" for 3D clocks, while ASUS has 400/900 in this position, there is something definately weird with these bioses. The only ones that work for me are ones using "Clockinfo3" as 3D clocks.

My card has 1.15v and 850/1200 in clock info 1. It's also stable @ that clocks and volts...but only is I use those infos in "clockinfo3" area


Maybe some cards are using a different controller...

It's 1.125V under load. And the original BIOS has 1.1625V according to RBE, it just doesn't listen to whatever you put in that register.

It seems some "cherry" cards that pass ATI's tests at 1.125V run at that voltage (let's call it "CPU registers A"), and don't look in the same spot as 1.1625V cards. So you can change RBE's CPU register voltage to whatever you want, the card won't listen as you are chaging "CPU registers B" and it's an "A" card.

That makes any sense?

RBE maker needs to find the spot of these extra registers so we can change it.

BTW afterburner of whatever can raise the voltage just fine.. (If you install afterburner and unlock voltage control it will be @ 1.125V volt. Pressing "reset to default" or whatever the button was called again sticks it @ 1.1625V, as Afterburner thinks that's the default voltage under load on all cards... Well obviously it isn't. And you can raise further just fine with Afterburner. I benched around a bit with it @ 1.3V and it gets much hotter so I guess that would mean it is actually raising the volts and not 'faking' it)

 

[cadaveca]

Well now..afterburner works fine for me with single card too, so it seems that MSI is aware of the difference, at least.

I mean, obviously my card here is different than yours, as the bios shows. So XFX is aware too.

Now I'm wondering...votls changes work for me, mem is about avg compared to other cards...if I end up RMA'ing my cards..what will I get?


So, like, do you have any other issues? Any GSOD's, 2DFlicker, or any of that?

 

[misterpro]

Well now..afterburner works fine for me with single card too, so it seems that MSI is aware of the difference, at least.

I mean, obviously my card here is different than yours, as the bios shows. So XFX is aware too.

Now I'm wondering...votls changes work for me, mem is about avg compared to other cards...if I end up RMA'ing my cards..what will I get?


So, like, do you have any other issues? Any GSOD's, 2DFlicker, or any of that?

On that card, one GSOD, withing 5mins of starting a 3D game. But ONLY on a cold system (as in powered off 24 hours). After that initial GSOD it can happily plow on for hours, days, weeks, whatever you want provided you don't shut down and power it off.

And it's nothing with the cooler either, I ditched the standard cooler.

Nothing with the Power Supply or drivers also. I put in a second 5870 (a 1.1625V one) I have without changing anything or reinstalling drivers. And it doesn't GSOD then ever...

Weird shit huh

 

[cadaveca]

Does the weird card has Hynix mem instead of Samsung? That might be part of it...

I was kinda shocked that the newer analogue parts came out so quickly too, so maybe ATI really is gonna release a whole new gpu this fall.

It's kinda a sad state of affairs when 7 cards don't work right @ stock clocks. Maybe all these differences is due to them dealing with it.

Anyway, I'm venturing into O/T here, so I'm gonna stop right there.

You've pulled the cooling, so you wanna maybe note the pcb number, etc? Maybe we can suss out the differences.

 

[misterpro]

Does the weird card has Hynix mem instead of Samsung? That might be part of it...

I was kinda shocked that the newer analogue parts came out so quickly too, so maybe ATI really is gonna release a whole new gpu this fall.

It's kinda a sad state of affairs when 7 cards don't work right @ stock clocks. Maybe all these differences is due to them dealing with it.

Anyway, I'm venturing into O/T here, so I'm gonna stop right there.

You've pulled the cooling, so you wanna maybe note the pcb number, etc? Maybe we can suss out the differences.

Where would that be located? On the back is a rectangle labeled

PLACE PERMANENT P/N BARCODE,
& S/N PRODUCT LABEL HERE

But there's nothing there Just like the pics on quite a few reviews. For example this techpowerup review pic

And it's using Samsung memory

SAMSUNG 934
K4G10325FE-HC04

 

[SoulsCollective]

All this sounds kinda odd. Did any of you guys actually measure (with a multimeter) what the voltages in fact are before/after RBE/Afterburner voltage setup?

Ah, the man himself.

No, I haven't verified voltages via DMM - I don't have a working one at present - but as above, we've done our best to verify Afterburner is working whereas RBE is not -

However, I can use Afterburner to set the voltage I want. In fact, testing different voltages with Afterburner is how I came up with 1.275v, since I found out my card can do 1050/1300 stable with 1.275v.

According to GPU-Z it is. And when raising with Afterburner I can see temps increasing on all relevant components, as well as being stable at higher clock speeds at which my machine will GraySOD/driver recover when at stock volts.

BTW afterburner of whatever can raise the voltage just fine.. (If you install afterburner and unlock voltage control it will be @ 1.125V volt. Pressing "reset to default" or whatever the button was called again sticks it @ 1.1625V, as Afterburner thinks that's the default voltage under load on all cards... Well obviously it isn't. And you can raise further just fine with Afterburner. I benched around a bit with it @ 1.3V and it gets much hotter so I guess that would mean it is actually raising the volts and not 'faking' it)

So, like, do you have any other issues? Any GSOD's, 2DFlicker, or any of that?

Nope, no issues whatsoever beyond the known issue of flicker when enabling Overdrive with two monitors. Card works perfectly.

 

[MakubeX]

Ah, I forgot about this thread. I'm glad other people brought it back.

Asus SmartDoctor (which what I'm currently using) can also change the voltages just fine. Just can't do it with RBE.

 

[MrLeopard]

Information regarding this problem is spread over several threads here. BAGZZlash said in another thread (http://forums.techpowerup.com/showthread.php?t=120022) that it's very unlikely that those 1.125V are "hidden" in a register that RBE can't read. As far as I know all "register 0x18 resistant cards" have those 1.125V 3D vGPU (measured with Afterburner/GPU-Z/Smart-Doctor) in common.

It's safe to say that overvolting via Afterburner/SmartDoctor does work on those cards, since overclocking-capability (and temperature) increases accordingly.

 

[MrLeopard]

Some news (kind of):

There is some guy at hardwareluxx.de who bought a second-hand HD5870 and the original owner flashed it successfully to 1000/1250 MHz @ 1.25V and used it this way. Anyway, the new owner found out that his new card is completely unstable and therefore monitored voltages during load: 1.125V

So if it's true that the original owner flashed and used the card with altered voltages and the new owner is affected by the 1.125V-phenomenon on the same card, this would mean that the cause of this problem lies somwhere in the remainig system. Maybe it's the mainboard/chipset, or psu, or operating-system?

 

[SoulsCollective]

Could be, but then, why would it specifically affect reading of one register and yet not Afterburner, Smart Doctor, etc etc? Is that even possible? Similarly, the problem seems quite widespread, and affects different motherboards, PSUs, OSes, etc etc.

Still, something to consider.

 

[MrLeopard]

Could be, but then, why would it specifically affect reading of one register and yet not Afterburner, Smart Doctor, etc etc? Is that even possible? Similarly, the problem seems quite widespread, and affects different motherboards, PSUs, OSes, etc etc.

Still, something to consider.

I'm afraid there are no other options left. Does it really affect different OSes, or is it all down to Win7? One could assume that nearly all HD5870 users are using Win7 due to DX11 support.
Maybe it's some kind of energy-saving-interaction between motherboard, os and driver in certain combinations? I'd like to see the behaviour of one those cards on a good-old WinXP 32-bit installation.

 

[BAGZZlash]

So if it's true that the original owner flashed and used the card with altered voltages and the new owner is affected by the 1.125V-phenomenon on the same card, this would mean that the cause of this problem lies somwhere in the remainig system. Maybe it's the mainboard/chipset, or psu, or operating-system?

Could be, but then, why would it specifically affect reading of one register and yet not Afterburner, Smart Doctor, etc etc?

Could be. Maybe it's similar to the fan control problem where the driver bullies the BIOS settings out as soon as the driver is loaded. TSR-Tweaking tools like AfterBurner again overwrite the settings in the memory afterwards and thus, still work. ATI seems to dislike BIOS modding and lets the driver overwrite the settings the BIOS puts in on bootup.

There are two ways to test this:

• First idea: Try a clearly older catalyst like 9.11 or something.
• Second (and even better) idea: Flash a BIOS with significantly more voltage than those 1.125V and measure it via multimeter during boot up. If it helps, enter the CMOS setup so the computer won't boot into windows. If the voltage is higher without loaded drivers and then, after booting, drops back to 1.125V, it's due to the driver.

 

[SoulsCollective]

Could be. Maybe it's similar to the fan control problem where the driver bullies the BIOS settings out as soon as the driver is loaded. TSR-Tweaking tools like AfterBurner again overwrite the settings in the memory afterwards and thus, still work. ATI seems to dislike BIOS modding and lets the driver overwrite the settings the BIOS puts in on bootup.

I'll do my best to test the first possibility as I don't have a DMM, but again, that raises a question - why would only some cards be affected? In some cases, cards from the same manufacturer of the same design with the same part number? I agree that this sounds most probable, but I don't understand how only some cards could be affected and not others.

 

[misterpro]

Could be. Maybe it's similar to the fan control problem where the driver bullies the BIOS settings out as soon as the driver is loaded. TSR-Tweaking tools like AfterBurner again overwrite the settings in the memory afterwards and thus, still work. ATI seems to dislike BIOS modding and lets the driver overwrite the settings the BIOS puts in on bootup.

There are two ways to test this:

• First idea: Try a clearly older catalyst like 9.11 or something.
• Second (and even better) idea: Flash a BIOS with significantly more voltage than those 1.125V and measure it via multimeter during boot up. If it helps, enter the CMOS setup so the computer won't boot into windows. If the voltage is higher without loaded drivers and then, after booting, drops back to 1.125V, it's due to the driver.

Intersting observation but why does my card run @ 1.125V while when I swap in my 2nd 5870 without changing anything or even reinstalling drivers it runs @ 1.1625V

That would suggest it isn't the driver either right?

 

[BAGZZlash]

Hmmm... true. Anyway: Isn't there anyone with a multimeter around?

 

[mkchiu]

[*]Second (and even better) idea: Flash a BIOS with significantly more voltage than those 1.125V and measure it via multimeter during boot up.
[/LIST]

The UPI power controllers I've measured (6204s) have "hard-coded" cold-boot (POST) voltages via resistor dividers. IIRC, the ST L6788A datasheets also indicate "selectable" hard-coded start-up voltages. I've no Volterra parts or datasheets.

So measuring POST voltage is not useful to verify RBE POST operation. Instead, it isoltes whether the PCB layout is poor, or GPU foundation BIOS from AMD is non-functional.

I've measured one 5770 with a ~1.176V and another 5770 with ~1.289V as the POST-start-up-voltage.

 

[MakubeX]

Yeah I don't think it's mobo or OS related because I used other 5870s of the same brand and this is the only one having the 1.125v problem.

 

[MrLeopard]

Yeah I don't think it's mobo or OS related because I used other 5870s of the same brand and this is the only one having the 1.125v problem.

OK, but there is at least one case where one single HD5870 features the 1.125V-phenomenon in one system and behaves like normal in the other system. Given that the person isn't talking trash of course...

 

[b2vFAh6J]

i'm getting the exact same problem. using XFX 5870 XXX edition, and flashed with asus bios as well, no change.

 

[SoulsCollective]

Any ideas on this?

I've picked up another 5870, again a Sapphire "Game Edition", same clocks, same reference design, except this one is not affected by the stuck voltage problem. Getting very frustrated not being able to properly OC one of my two, apparently identical, cards.

 

[Enoy21]

I think I may have found something here guys.

I don't believe it's a hardware thing but a Bios thing..... And could be related to the card itself.

Those of you using two 5870 cards with different results.... Are you SURE it's truly two 5870's and one is not a 5850 ? Are the majority of us in here trying to flash a 5850 to 5870 bios having this issue ?


I am running a 5850 which uses 0x17 as it's 3d performance register.
The 5870 bios (and actual card possibly) uses 0x18.

If I install the 5870 Bios I have the same issue as all of you I top out at 1.125 no matter what setting is in 0x17 or 0x18 ( I tried setting them both to 1.1625 ) But I get my VTT Register controls.


If I put on my 5850 Bios I can set the 0x17 to whatever I want and it sticks and shows in GPU-z. I used 1.15 as my 0x17 register and it worked. But I lose my VTT controls.

One other thing I noticed is that on the 5870, the Clock 00 shows a voltage of -- rather than what it's reading in the registers, whereas the 5850 shows the register information in clock 00 as what is found in 0x17 . ( in this case 1.088 )


I'm running a test right now and took my 5850 Bios. Changed the register 0x17 to 1.1625 ( 5870 voltage ) BUT I changed the clock 00 voltage to --- which is what the 5870 shows.

After reboot I'll be back for update.


====================


Ok after setting the Clock00 to --- voltage it seems to have defaulted to 0x15 (1.0v) at idle so that's screwing with my powerplay options.


So right now it seems I have one of two options.


1 - Use a 5870 Bios and get higher memory OC thanks to the higher vMem but lower GPU due to lack of vCore options
2- Or the other way around. Use a 5850Bios and whatever vCore I want but less vMem thereby reducing Memory OC.


Unless someone has come up with a better answer since this time one year ago ?

 

[Enoy21]

Could this actually be a problem with GPU-Z reading it correctly ?

It seems Furmark and GPU Shark are reading what the Bios says it should be doing :

 

[Enoy21]

Since I am just talking to myself here I'll update myself.


I am believing this is just an issue with GPU-z as I am currently benchmarking a 925-1250 OC. And the above picture shows the two other programs reading differently.

I believe that going to 1.2v core and slightly higher on memory I could hit the 1000/1300 mark , but I don't think that is necassary at this time as I don't want to have to run my fan above 40% due to noise levels.

With stock cooling ( reference ) and auto fan turned on I am topping out around 86c in Furmark , but my games and Unigine don't seem to go above 75c .


One change I also made today that may have helped in stability ( can't say for sure ) is that I went to the ATI 103 bios rather than the Sapphire/Asus 101 I had been using previously.

This increased vcore (register 0x18 ) voltage from 1.165 to 1.175.


I am so glad I took the chance and used RBE instead of third party active software due to the punkbuster interference. ( If not for Punkbuster then any of the overvolt Utilities would be awesome.)

Now I can OC with ATITray tools with no punkbuster issues at all.

Unigine scores went from 43 FPS score 1103 @ stock to 56 fps 1364 score at 925/1250. ( although there was no real changes between 1225 and 1250 on memory so I will likely run 1225 for most applications )


Thanks for reading. Maybe this will help someone else in a google search one day.