7870 boost harmful voltage?

[CookieTrooper]

Hi everybody!

I recently got a Sapphire 7870 Boost (the one with a Tahiti LE Chip). I wanted to undervolt it, so I monitored it's voltage with GPU-Z. With the Sapphire Trixx Tool I lowered the Voltage from 1187 to 1000 without running into stability issues but got some strange data.

With the undervolted setting I got short voltage peaks, up to 1.470V, which i think could be dangerous for the chip. I ran this test with Furmark and Tomb Raider multiple times, so if the reading is wrong, it's very consistant.

When I tried to run Furmark at a fixed voltage setting of 1000 (which ran fine without "fixing" it), Furmark crashed.

When i tried Minecraft with the stock voltage i saw the same results as with the undervolted setting in Furmark and TR - short but very high voltage peaks.

So here are my questions:

- Are nearly 1.5V over a short period of time dangerous for the lifespan of a Tahiti LE GPU?

- Is there anything I can do about these peaks except setting the voltage to a fixed level?

- Is it possible that these peaks are just an error in measurement?

Thank you very much!

 

[tokyoduong]

I'm sure that is what the card is supposed to do. I'm at work so i can't look up all the info but it looks like the card can dynamically throttle/boost itself based on TDP headroom.

AMD POWERTUNE AND ZEROCORETECHNOLOGY
Maximizes performance under load conditions by dynamically increasing the GPU engine clock to take advantage of unused TDP headroom. Also allows users to configure their own TDP limit, within a provided range, for even higher performance or more power efficiency. AMD ZeroCore Power technology ensures ultra-low idle power when the connected display is in power saving mode.

 

[erocker]

Try using a different monitoring program for voltage such as Afterburner or HWinfo and see what they say.

 

[d1nky]

afterburner shows target voltages, mod the ven file in afterburner to display 'true' voltage

http://forums.overclockers.co.uk/showthread.php?t=18431335

 

[CookieTrooper]

Followed the Instructions, got the same results:

When undervolted the card peaks at 1.44V when Furmark starts up, and jumps right back to 0.95V

 

[erocker]

Strange issue for sure. Try contacting Sapphire support and see what they have to say: https://support.sapphiretech.com/presentation/member/member_login.aspx

 

[d1nky]

could this be like reverse vdroop? ive seen it mentioned about cpus, cant remember where or when.

maybe a dodgy vrm, can you see the vrm breakdown and temps? hwinfo will display it if the card has sensors for it

When undervolted the card peaks at 1.44V when Furmark starts up, and jumps right back to 0.95V

that looks like LLC to me, counteracting power draw with more voltage. but does the card have LLC features.....baffled...

 

[CookieTrooper]

could this be like reverse vdroop? ive seen it mentioned about cpus, cant remember where or when.

maybe a dodgy vrm, can you see the vrm breakdown and temps? hwinfo will display it if the card has sensors for it



that looks like LLC to me, counteracting power draw with more voltage. but does the card have LLC features.....baffled...

without undervolting the vrm temp gets realy hot - nearly 80°C. one of the reasons why i wanted to undervolt in the first place. however, the voltage peaks seam not related, they occur with hot and cold vrm. furthermore, when undervolted, the vrm stay cooler, yet i get more voltage peeks

I'm sorry but i don't know what LLC means (google and wiki weren't too helpful)

 

[d1nky]

its designed to counteract the voltage drop when theres an increase in powerdraw LLC. so delivers more voltage when power is drawn from the source. (its called loadline calibration)

I would phone the retailer or even put a ticket in for RMA because the vrm temps and voltage spikes (im not sure about your card but mine stays under 60*c in a hot room, and just go above 60 when ocd to death)

im not sure if youre allowed to adjust voltages etc if that voids warranty, youll have to research.

maybe keep an eye on ambient temps, fan speed etc. and what is the core temp in contrast to the vrm temp?

 

[erocker]

This is why I mention to contact Sapphire support. Perhaps a bios they give you will fix it. Perhaps not, the card could just be defective.

 

[CookieTrooper]

its designed to counteract the voltage drop when theres an increase in powerdraw LLC. so delivers more voltage when power is drawn from the source. (its called loadline calibration)

that seams highly likely, since furmark crashes when i fix the voltage on otherwise stable levels.

I would phone the retailer or even put a ticket in for RMA because the vrm temps and voltage spikes (im not sure about your card but mine stays under 60*c in a hot room, and just go above 60 when ocd to death)

just to clarify this: do you have a 7870 boost? are we talking about the temps GPU-Z shows as "VRM Temperature 1&2"?

im not sure if youre allowed to adjust voltages etc if that voids warranty, youll have to research.

pretty sure it does void warranty, but honestly i'm not looking for a replacement as long it works

maybe keep an eye on ambient temps, fan speed etc. and what is the core temp in contrast to the vrm temp?

ambiant temps are quiet low atm (damn winter, go away!), but i'm using an isolated lianli miditower which tends to get quite hot. with an open case vrm temps are significantly lower.

This is why I mention to contact Sapphire support. Perhaps a bios they give you will fix it. Perhaps not, the card could just be defective.

yeah i'll contact them if i can't fix it myself (with some help of yours ), i don't have the best experience with support from any company, so this would be the last option

 

[d1nky]

when contacting them probably best to miss out the details of modyfing voltages lol

I use hwinfo and I have asus 7950. and can see vrms, their voltage breakdown and currents etc. (use hwinfo64 and see if there is any strange voltages, could be defective like mentioned before)

research LLC, see if the card has this feature. research the cards typical vrm temps. I know vrms can get toasty and 80*c may be normal?! and you didn't say about the graphics core temp?!

could just be a combination of factors contributing to higher temps and you undervolting is making it unstable so the card spikes the voltage up to load the power draw and then resets.

also you said (and I didn't pick up on it) ''when Furmark starts up, and jumps right back to 0.95V''

0.95v seems low to run full load when normal usage voltage is 1187. vdroop included maybe youre causing the voltage spike due to lack of voltage.

try running stock voltage and bump up fan speeds and keep temps low. see if there is a spike? if yes then contact support, if no then stop undervolting and research a bit

 

[CookieTrooper]

and you didn't say about the graphics core temp?!

the gpu stays below 69°C no matter how long furmark runs. with the case opened it stays below 65°C

could just be a combination of factors contributing to higher temps and you undervolting is making it unstable so the card spikes the voltage up to load the power draw and then resets.

i've rerun all my little tests with the case opened, results are exectly the same. temperature really shouldn't be the cause.

0.95v seems low to run full load when normal usage voltage is 1187. vdroop included maybe youre causing the voltage spike due to lack of voltage.

thats quite the purpose of undervolting, isn't it? but as i stated before, i get these voltage peaks with stock voltage too. not in furmark or tomb raider but strangely enough in minecraft.

try running stock voltage and bump up fan speeds and keep temps low. see if there is a spike? if yes then contact support, if no then stop undervolting and research a bit

as i said, yes, there are. i'll contact support monday and let you guys know what they have to say. i stopped undervolting as soon as i saw the spikes i researched quite a bit, even posted in another forum 2 weeks ago but the guys over there didn't know either...

 

[Domestic_ginger]

Suggestion!

Overclcok to the top of the boost range (maybe +1mhz); this should stop boost kicking in and see what happens. AFAIK boost stops when you go above the boost clock.

I was interested in getting one of these (well a myst one) and was hoping for successful undervolting to follow.

OMG; one month late.

 

[CookieTrooper]

Sry for not answering, havn't checkted my mails for a while. The problem with your sugestion is that Afterburner doesn't let you overclock the basefrequency, only the boostfrequenzy. But your replay inspired me to try it with the power limit set to +20 in order to keep the frequency steady.

This fixes the problem with the voltage peaks. FALSE!

Sadly, it also prevents undervolting, the card stays at 1.170V no mather what the CoreVoltage is set to (at least in Furmark). Without 3D workload the voltage goes down as expected.

EDIT: This only fixes the voltage peaks in Furmark, Minecraft doesn't care.

 

[Jstn7477]

I get erroneous readings sometimes with my cards e.g. vcore current spikes upwards of 200A without corresponding voltage droop when normal current use is 80-130A. The voltage controller's monitoring isn't going to be anything close to a good digital multimeter or oscilloscope, so unless you are going to employ such tools to physically measure your VRM output, it's probably just a monitoring anomaly.

My 7970 running F@H, a load that doesn't tend to vary unless it pauses for a few seconds. Notice the current drops and spike without corresponding voltage changes. The only corresponding voltage and current change is when the load briefly pauses. All the other times, the voltage doesn't experience noticeable changes yet the current is all over the place. Yes, I know I left 12V and VDDC set to MAX in the screenshot, but the VDDC stays at a continuous 1.129-1.131v under this particular load.

 

[CookieTrooper]

Thanks for replying, i have given up hope of finding a solution some time ago

Yeah, there is always potential for error in measurement, sadly i don't have the equipment to check it external.

On the other hand, there clearly is a pattern for these peaks. They only occur when the card is running below the maximum voltage set. But instead of setting itself to the maximum voltige, the card overshoots and reaches annoying high voltage peaks.

Because of the fact that this peaks don't seam to have any effect on the performance of the card there is really only one question: Are 1.45V for a short period of time, but again and again harmful for the lifespan of the card? If not i would like to resume undervolting

 

[qubit]

@CookieTrooper

It sounds to me like these voltage spikes are the card compensating for too low a voltage to start with. It makes sense that when a large percentage of the chip is doing work it'll need extra oomph to work without crashing and that's what those voltage spikes are doing, I believe.

I think you'd be better off leaving the voltage at stock, or if you really want to lower it, lower the GPU clock, too and see how it behaves. Of course, you'll lose performance, but it's the price you pay for running the card out of spec.

Note for comparison, that the higher performance, faster clocked CPUs from Intel tend to have higher operating voltages and TDP, so there's no getting around it.

So, could those big voltage spikes hurt it? Probably not, but as it's not what it does when run at stock, you can take a hint from it that this low voltage strategy probably isn't that good for it.

Hope this helps.

 

[CookieTrooper]

So, could those big voltage spikes hurt it? Probably not, but as it's not what it does when run at stock, you can take a hint from it that this low voltage strategy probably isn't that good for it.

Well, as I said, I do get these strange peaks with stock voltage too...

 

[Domestic_ginger]

They've released a 7870 xt version 2.0. Maybe worth looking into; this card is sold by OCUK and I can find no other details for it even on Sapphires website.

I got a Myst and at 1050mV the card is stable. 1000 it appears to be unstable but I only got 58.9% asic. at load I get droop down to about 1020 according to GPUz. Powercolours Powertune will change the base clock speed (the voltage tuning does not work; its pretty basic). I think HIS's iturbo will too but the lowest voltage range is about 1167.

I'm using trixx and keeping stock voltage. i would use afterburner but the profiles appear to be disabled on the latest Beta.

 

[qubit]

Well, as I said, I do get these strange peaks with stock voltage too...

Then it's normal for the card.

 

[CookieTrooper]

They've released a 7870 xt version 2.0. Maybe worth looking into; this card is sold by OCUK and I can find no other details for it even on Sapphires website.

My guess would be that this is a normal Sapphire 7870 XT Boost edition. The rev2.0 only means that the card isn't the original 7870, which is true. Why? Otherwiese there would be a test of some sort somewhere (at least i didn't find anything). Also the card was mentiont 4 months ago ( http://uk.answers.yahoo.com/question/index?qid=20130203204421AAzQHWJ ), so it doesn't seam to be a refresh.

I got a Myst and at 1050mV the card is stable. 1000 it appears to be unstable but I only got 58.9% asic. at load I get droop down to about 1020 according to GPUz. Powercolours Powertune will change the base clock speed (the voltage tuning does not work; its pretty basic). I think HIS's iturbo will too but the lowest voltage range is about 1167.

Did you notice any volatge abnormalitys with your card?

Then it's normal for the card.

Most likely you are right. I really should stop worrying about such litte things^^

EDIT: Here is the proove that the OCUK model is just a normal 7870 Boost: Search the manufacturers code in google and you get the normal version.

 

[qubit]

Most likely you are right. I really should stop worrying about such litte things^^

No sweat buddy, I can totally relate to concerns like this.

 

[Domestic_ginger]

Did you notice any volatge abnormalitys with your card?

Nope. I get 1.256V spikes when it boosts if I don't lock the voltage but nothing as high as you're getting.

Boost is wrong and boost and powertune is doubly wrong. I wish they would just let you turn them off.

I would have a look at your bios version and see what everyone else is running tbh.