18 replies to this topic

[Texugo87]

I finally got my power supply back on warranty, so I've been working on dialing my overclock back in. Unfortunately my graphics card is giving me some issues.

I've been using MSI Afterburner and my system specs are as follows:

i5 3570K Oc'd to 4.4ghz
Asus p8z77 LE Mobo
Power Color HD7870 LE Myst
8 gigs Mushkin DDR3-1600
128gig SSD
500gig HDD
550watt Seasonic Powersupply (Single rail with 45amp capacity)

The 7870 has a stock clock speed of 975mhz, which is apparently actually 925mhz with a "boost" to 975mhz when needed. I started off with a gentle OC to 1025mhz, and increased the power limit in afterburner, which I knew was previously a perfectly workable OC, I can't remember where I OC'd it to before but I think it was around 1200mhz before I experienced artifacting.

Anyway, I fire up MWO and leave MSI running in the back ground. Things run fine for the first couple minutes, FPS almost pegged on 60fps. Then I start getting stuttering and FPS dips into the 40s and 30s. After the match I look at my afterburner log file, and basically what was happening was when the stuttering and dipping started happening, my gpu utilization was spiking up to 95%-100% and core clock was dipping from 1025 to 972mhz. Temperatures were fine.

Unfortunately, I had seen this behavior before, after my PS tanked, and I put in my backup PS, an ancient supposedly 600w supply with 2 rails with only 18amps each. I had figured it couldn't keep up with my system requirements, and this was the impetus for me reverting my GPU and CPU clocks to stock until I got my PS back on warranty.

After that I reset the clock speed to stock, which MSI registers as 975mhz and fired MWO up again leaving my settings the same as the previous attempt. This time both MSI and GPU-z reported my clock speed as a stable 925mhz, and the game ran like poop (relatively), with FPS down in the 40's most of the time. I decided to call it a night.

This morning I fire everything up again, and MSI and GPU-z both report 975mhz. At first I get the kind of FPS i'm used to, then again comes the dips and stuttering (not particularly related to anything happening ingame). I alt tab to MSI and see again, GPU utilization going from a stable 70-80% to peaking at 100% while the core clock simultaneously crashes to 925mhz, then utlization dropping of and clock speed picking back up, then the same thing repeating again.

Voltage is stable, temperatures are stable in the high 60's. I'm at a loss for what is causing this. I'm about to stress test in furmark and see if I can get the same behavior.

funny enough 975:925 and 1025:972 are about the same ratio.

Any suggestions?

TLDR;

My GPU clock speed simultaneously tanks and spikes in sync with in-game stuttering, at the same time the GPU clock falls my utilization spikes to 100%. No thermal, or power problems. Help?

Edited by cjmurphy87, 11 December 2014 - 08:06 AM.

[Texugo87]

Update: Ran it in furmark and it wasn't pretty, and the clock speed gets lower and lower over time, at one point it was spiking between 975mhz, and 573.5mhz.

Luckily, my card has about a month and a half left on the warranty.

[xWiredx]

Could definitely be the card. If it were out of warranty, I would start making recommendations. Since it isn't, though, might as well see what the manufacturer can do for you instead.

There is also the more remote possibility that it is the mobo (PCIE slot). I had a buddy run into this situation and it happened for a few months before he got tired of it. One day he just no longer got video output and thought it was his card, but it ran fine in my system. Putting it in the second PCIE slot on his mobo worked great until he built a new system a few months later. If the manufacturer gets you a new card, and you still experience the same issue, you might want to check the card in another known-good motherboard.

[Texugo87]

Thanks for the idea, I'll toss it in my second PCI-e slot just to be sure while I'm waiting for the manufacturer's tech support line to open for the day.

[Flapdrol]

How hot does the gpu get?

[Texugo87]

65-70*c.

[Flapdrol]

Hm, that shouldn't cause throttling, maybe the vrms get hot or something, does it do it if you force the fans to 100%?

I'd test with unigine heaven instead of furmark by the way.

[Texugo87]

I'll give it a shot in unigine. I just ran it through furmark in the second GPU slot, and set the fan to 100% out of curiosity, I got clock instability within about 5 seconds of testing, temperature was below 45*c. Core voltage is stable, so if the Vrms were the problem I would expect to see fluctuations, right?

[Texugo87]

Unigine showed the same fluctuations as MWO, mostly rocketing between 925 and 975, with occasional dips into the 800's. My memory clock this go around logged as 1361 throughout when it should be 1500mhz...

It would clock 975 at about 65% load, then go to 925 with 100% load, and back again repeatedly, temperature was 60-63*c.

Edited by cjmurphy87, 11 December 2014 - 08:55 AM.

[Flapdrol]

cjmurphy87, on 11 December 2014 - 08:35 AM, said:

I'll give it a shot in unigine. I just ran it through furmark in the second GPU slot, and set the fan to 100% out of curiosity, I got clock instability within about 5 seconds of testing, temperature was below 45*c. Core voltage is stable, so if the Vrms were the problem I would expect to see fluctuations, right?

Throttling between 925 and 975 is normal behaviour, happens when the card goes over the allowed power limit, with overheating vrm's (or gpu itself) you'd get more massive throttling, like the 575 MHz you got with furmark.

[Texugo87]

Thanks,

That's basically what PowerColor's Tech support depart said. They said this is normal for these cards as the power supply components age and would not be something covered under warranty; a suggested solution was increasing the power limit in MSI or equivalent until it stabilizes, which I've now bumped up a bit more and am seeing stable clock speeds in both unigine and even furmark at 975mhz.

Further update: Just played around of MWO and the stuttering and diping in frame rate are gone, performance is back to what I'm used to, logs showed a solid 975mhz through match, though memory clock is still reading 1361 instead of 1500 for some reason.

Edited by cjmurphy87, 11 December 2014 - 09:36 AM.

[Texugo87]

Final update: increasing the power limit resulted in an almost 20% increase in unigine heaven score, which is now inline with the score I was getting when I first got this card again. I have significant doubts as to the long term out look for this card though, and am not entirely impressed that the manufacturer considered this rate of degradation normal.

[xWiredx]

I've never had any videocard from any manufacturer I've ever purchased degrade that quickly, nor have I had support people tell me that it was normal. My Radeon 9600XT worked roughly the same as it did brand new up until about 3 years ago when the machine it was in was taken out of service. I still have a Geforce 9800GTX+ that benchmarks about the same as it did brand new, too. Even looking at cards from around the same time as yours came out, I have a Geforce GT 530 and 2 GTX 660 TIs that are both working like brand new. The GT 530 only ever ran CS:S and Left 4 Dead, but it was OCed about as far as it could go. The 660 TIs are both heavily factory OCed. I would be giving this thing up on warranty for severe degredation. No way are cards supposed to decline that quickly.

[Superlag]

I had the exact same issue with my new 7970 graphics card from Sapphire. I used both MSI afterburner and Sapphire Trixxx overclocking applications but I still got the stuttering issues. I fixed these issues by actually enabling and overclocking the card via Catalyst Control Center. I set the core and memory clocks to the desired speed and set the power limit setting to +20. Doing so this way allowed the clocks to stay stable and eliminate the studdering issue. Although it doesnt allow you to modify the voltage rates I believe you can then edit the card further via the aftermarket applications (Set CCC overclock settings to card's stock settings and power limit option to +20 then use MSI/Sapphire app to overclock/overvolt further).

I have not tried it but you might also want to prevent the Catalyst Control Center application/services from loading so only your 3rd party overclocking utility is active. I've also noticed that Catalyst Driver version 14.4 is more stable with these HD 7xxx series cards. BTW this issue is caused by the GPU being signaled to be in 2D/low power mode forcing the clockrate to spike up and down as it doesnt know which state to be in, 2D or 3D. This is a known issue with most newer AMD video cards.

Also, when studdering, artifacts, or weird performance problems occur you must restart your machine for these problems to go away.

P.S. - Catalyst driver recommendation may only apply to those running Eyefinity setups as I have not tested with a single monitor.

Edited by Superlag, 11 December 2014 - 11:58 AM.

[Flapdrol]

cjmurphy87, on 11 December 2014 - 09:24 AM, said:

That's basically what PowerColor's Tech support depart said. They said this is normal for these cards as the power supply components age and would not be something covered under warranty

That's insane, the base clock of these cards is 925, it should only go lower in emergency cases, like overheat. If it's "normal" in actual games in a well ventilated case powercolor is a terrible graphics card designer.

Maybe if there's dust on the vrm's, and that's why they get hot, but otherwise a fault in the design of the card.

[Texugo87]

Flapdrol, on 11 December 2014 - 12:22 PM, said:

That's insane, the base clock of these cards is 925, it should only go lower in emergency cases, like overheat. If it's "normal" in actual games in a well ventilated case powercolor is a terrible graphics card designer.

Maybe if there's dust on the vrm's, and that's why they get hot, but otherwise a fault in the design of the card.

No dust, I gave it a good cleaning out when I tried it in the second PCI slot.

Superlag, it looks like I have to set the power limit at +12% to get it to hold a constant clock setting in MWO.

It's worth noting that it did not flip between clock settings like this when new. Not at stock settings, not overclocked, not when doing a stress test in Furmark.

Yeah, the representative basically told me that the frame rate stuttering (which i was experiencing in other applications as well), was considered normal as well... I think I'll call back in a coupe days and see what another representative has to say. I'm hesitant to spend money shipping it back to them, and potentially return shipping if they refuse the warranty, instead of ear marking it for a new card eventually. Just to clarify once I upped the logging freqency in Afterburner to every 200ms, I was seeing dips into the 800hmz range at stock power settings, this too the respresentive said was "normal" for these cards.

This is my first AMD card (well, second as the first 7870 I got went ****-up after 1 month), before that I had a GTS 250 that worked flawlessly for years, and my living room computer has a 10 year old Nvidia card of some sort..

[Durant Carlyle]

All electronic components can and do lose performance as they age. Sometimes they lose performance all at once (total failure) and sometimes they lose performance slowly (like what the OP is describing).

Both are normal.

Not "insane" or unusual. It's just the way computer components are.

[Flapdrol]

Durant Carlyle, on 11 December 2014 - 01:55 PM, said:

All electronic components can and do lose performance as they age. Sometimes they lose performance all at once (total failure) and sometimes they lose performance slowly (like what the OP is describing).

Both are normal.

Not "insane" or unusual. It's just the way computer components are.

It shouldn't degrade to the point of emergency throttling within 2 years (or whatever the warranty is)

[MechWarrior4172571]

cjmurphy87, on 11 December 2014 - 06:32 AM, said:

I finally got my power supply back on warranty, so I've been working on dialing my overclock back in. Unfortunately my graphics card is giving me some issues.

I've been using MSI Afterburner and my system specs are as follows:

i5 3570K Oc'd to 4.4ghz
Asus p8z77 LE Mobo
Power Color HD7870 LE Myst
8 gigs Mushkin DDR3-1600
128gig SSD
500gig HDD
550watt Seasonic Powersupply (Single rail with 45amp capacity)

The 7870 has a stock clock speed of 975mhz, which is apparently actually 925mhz with a "boost" to 975mhz when needed. I started off with a gentle OC to 1025mhz, and increased the power limit in afterburner, which I knew was previously a perfectly workable OC, I can't remember where I OC'd it to before but I think it was around 1200mhz before I experienced artifacting.

Anyway, I fire up MWO and leave MSI running in the back ground. Things run fine for the first couple minutes, FPS almost pegged on 60fps. Then I start getting stuttering and FPS dips into the 40s and 30s. After the match I look at my afterburner log file, and basically what was happening was when the stuttering and dipping started happening, my gpu utilization was spiking up to 95%-100% and core clock was dipping from 1025 to 972mhz. Temperatures were fine.

Unfortunately, I had seen this behavior before, after my PS tanked, and I put in my backup PS, an ancient supposedly 600w supply with 2 rails with only 18amps each. I had figured it couldn't keep up with my system requirements, and this was the impetus for me reverting my GPU and CPU clocks to stock until I got my PS back on warranty.

After that I reset the clock speed to stock, which MSI registers as 975mhz and fired MWO up again leaving my settings the same as the previous attempt. This time both MSI and GPU-z reported my clock speed as a stable 925mhz, and the game ran like poop (relatively), with FPS down in the 40's most of the time. I decided to call it a night.

This morning I fire everything up again, and MSI and GPU-z both report 975mhz. At first I get the kind of FPS i'm used to, then again comes the dips and stuttering (not particularly related to anything happening ingame). I alt tab to MSI and see again, GPU utilization going from a stable 70-80% to peaking at 100% while the core clock simultaneously crashes to 925mhz, then utlization dropping of and clock speed picking back up, then the same thing repeating again.

Voltage is stable, temperatures are stable in the high 60's. I'm at a loss for what is causing this. I'm about to stress test in furmark and see if I can get the same behavior.

funny enough 975:925 and 1025:972 are about the same ratio.

Any suggestions?

TLDR;

My GPU clock speed simultaneously tanks and spikes in sync with in-game stuttering, at the same time the GPU clock falls my utilization spikes to 100%. No thermal, or power problems. Help?

cjmurphy87, on 11 December 2014 - 08:35 AM, said:

I'll give it a shot in unigine. I just ran it through furmark in the second GPU slot, and set the fan to 100% out of curiosity, I got clock instability within about 5 seconds of testing, temperature was below 45*c. Core voltage is stable, so if the Vrms were the problem I would expect to see fluctuations, right?

I would suggest putting the video card on a different rail of the PSU or trying a different PSU and I would also suggest (if you keep the card instead of sending it out for warranty) for you to DIY replace the thermal paste that the manufacturer put in--it is probably dry by now and that could account for some instability/loss of performance. I would strongly suggest you try different representatives at different times and don't mention first representative's assessment at all to other representatives. I think you should just RMA this card if it throttles for no apparent reason like that. As far as component degradation--2 years is WAY way too low for that. My MSI GTX 970 is boasting components that are stated to not degrade for 10 years. Most of the degradation can come from electrolytic capacitors and they put good capacitors in nowadays that last longer than 2 years of non-stop usage. I had to RMA an MSI video card before (ended up being something else that was causing the problem anyway) and they honored the warranty and gave me a refund.. was very pleased with their RMA service, and I didn't talk to no representatives--just RMAed it (for the problem I thought it was causing--ended up being a bad Dell monitor.) AMD processors are notoriously hotter and are clocked aggressively with voltage boost to make them competitive with NVidia.. so, I would be surprised if the thermal paste is not drying up in your card.