Help - Sapphire 7950 issues

[carlboyer]

Okay, first off, sorry for the length of this post, but I want to ensure all appropriate info is here. I have a Sapphire HD 7950 3GB OC ( 11196-02-40G) base factory OC clock of 900 MHz that keeps freezing on me - screen locks up either all one color or vertically striped. There are certain conditions that it tends to consistently freeze under - after or during Youtube videos and during the Windows Experience Index test. Other times it seems to randomly freeze. On two occasions it made it through the Index test :\. Running videos (WMV, MPEG, MP4, AVI, etc) from disk does not cause an issue. It does not freeze running Furmark.

I've run an NVidia GTX 460 and Gigabyte 1GB 6950 in my rig to test without any issues. A friend even took my card to run in his system for a weekend and it ran without issues in his rig. In fact, he was very impressed with the card. I've run Windows memory testing on my rig and that came back clean. I did find that the memory timing was wrong (XMP had not been enabled) on my motherboard, but even after correcting that the freezing continues. I also tried different combinations of my memory sticks and still it freezes. I have not noticed anything out of the ordinary while watching HardwareMonitor and doing testing. The freezes seem to come out of nowhere. It is just strange. Sapphire is suggesting trying a bit of overclocking my system as my bud's system is overclocked. Looking for suggestions/help before I start monkeying with over or under-clocking. Thanks in advance for any help. Details on my system and my friend's follow:

Graphics Card
- SAPPHIRE Radeon HD 7950 OC 3GB 384-bit GDDR5 PCI Express 3.0 x16 HDCP Ready CrossFireX Support Video Card ( 11196-02-40G) – Factory overclock.

AMD Catalyst Driver Version, and Driver History- Current 13.4, tested 13.1, 12.2, 12.4, 12.8

Operating System
- Windows 7 Home Premium 64 bit SP 1

Issue Details
- Video card randomly freezes under varying operating conditions. Most typically when 75% through Windows Performance Index test ( Tuning Windows Media Decoding) and right after viewing Yahoo or Youtube videos. Other times seemingly at random. Because it is just a freeze, there is no memory dump.

Motherboard or System Make & Model
- GIGABYTE GA-Z68A-D3H-B3 rev 1.3 with firmware F13

Power Supply
- CORSAIR Professional Series Gold AX850 (CMPSU-850AX) 850W
- http://www.newegg.com/Product/Produc...82E16817139015

Display Device(s) and Connection(s) Used
- Dell 1908FP with DVI connection

Applications and Games where issue seen occuring
- Windows Control Panel & Device Manager
- Yahoo and Youtube videos using IE 10
- Windows Performance Index Test
- Gigabyte EasyTune 6 launch (occasionally)

CPU Details
- Intel i7-2700K, stock settings

System Memory Type & Amount
- G.SKILL Ripjaws Series 16GB (4 x 4GB) DDR3 1600 (PC3 12800)

Additional Hardware
- Mushkin Enhanced Chronos MKNSSDCR60GB 2.5" 60GB SATA III SSD
- WD Green WD7500AARX 750GB IntelliPower 64MB Cache 6.0Gb/s SATA HD
- HGST Deskstar 7K1000.C 0F10383 1TB 7200 RPM 32MB Cache SATA 3.0Gb/s 3.5" HD

Buddy's system specs:
Intel i7 2600k OC 4.6ghz
Asus p8-z68 mobo
16gb Ram Gskill Ripjaw ddr3 12800
OCZ Agility 3 120gb SSD
2x western digital Black 2.5tb
Windows 7 x64
Antec 900 case
Antec true blue 750watt ps

* System moderately overclocked so running higher voltages.

 

[Jaydip]

Just run with a single ram stick and only the bootable hdd.Run any benchmark like Heaven 4.0 and report back.

 

[chadashcroft22]

Screen lockups with colors like that is vga card 100%. The only thing you can really do is try to underclock the card and see if it runs stable. But even if it does most people wouldn't be happy with it so I would just rma it because it will just get worse with time. And it's most likely the card ram that can't handle the clocks.

 

[MTDEW]

Since it works just fine in another system, I personally would....

Grab a free drive imaging program and make an image of your main OS drive as it is now.
You have the space to store the image on your other drives and with an SSD, creating and restoring an image would be quick and painless.

Then....
1: Unplug all extra hard drives and anything else that isnt needed.
ie: just mouse , keyboard , main OS drive....only what is needed.
No overclocking anything.

2: Then i'd do a clean install of windows with just all windows updates and mobo hardware drivers. (no other programs installed)

3: Install latest GPU drivers and test.
I'd test overall system stability with memtest, prime 95 etc... as well as the GPU testing while i had a "clean" install of windows to work with.
(Well, windows isn't needed for memtest, but you get the idea.)

 

[carlboyer]

Thanks for the suggestions. Here is an oddity I'm seeing, though. In the bios, after turning on XMP and selecting Profile 1, the system loaded the following memory settings:
tcl = 9
trcd = 9
trp = 9
trp = 24
Command rate = 2
However, the MB Intelligent Tweaker shows tcmd = 1. Does that mean my command rate has been reset to 1? I went in and manually reset these setting and tcmd came back at 2 for two reboots, but not keeps coming back 1. I'm assuming tcmd = Command Rate.

 

[boboftw]

Right click a youtube video and disable hardware acceleration. See if that helps.

 

[carlboyer]

Well, I did a clean install on my SSD and it on my first try at running the Windows Experience Index the card froze.

Since it works just fine in another system, I personally would....

Grab a free drive imaging program and make an image of your main OS drive as it is now.
You have the space to store the image on your other drives and with an SSD, creating and restoring an image would be quick and painless.

Then....
1: Unplug all extra hard drives and anything else that isnt needed.
ie: just mouse , keyboard , main OS drive....only what is needed.
No overclocking anything.

2: Then i'd do a clean install of windows with just all windows updates and mobo hardware drivers. (no other programs installed)

3: Install latest GPU drivers and test.
I'd test overall system stability with memtest, prime 95 etc... as well as the GPU testing while i had a "clean" install of windows to work with.
(Well, windows isn't needed for memtest, but you get the idea.)

 

[Piotrsama]

Have you tried with only 1 stick of memory?
Also, try memory at 1333 Mhz

 

[MTDEW]

Well, I did a clean install on my SSD and it on my first try at running the Windows Experience Index the card froze.

Good, mission accomplished.
The idea was to eliminate any possible software issues.
Did you also try something like prime 95 testing to see if it freezes on something non-GPU related?
ie: Its usually really hard to tell if its the card or the system "freezing".

If so, I'd say its time to look at hardware/bios settings.
You mentioned your memory, there's no harm it testing it at 2T just to "be sure".
(not sure if it's actually rated at 2T or not, you weren't clear either way)
I'd manually set it at MFR rated speed/timings and voltage and test each stick individually, then all together.

I know running memtest is boring and time consuming, but it certainly looks like your best place to start.
And testing one stick at a time can get old fast.
But at this point, I'd just go ahead and do it just to be sure so you can eliminate it as a possible issue.

 

[carlboyer]

Yeah, I'm figuring I'm headed for memory testing - ugh. The MFR recommended timing on the memory is 9-9-9-24-2N. Not sure what the N is for - haven't figured that out. With XMP on using Profile1, those are the settings, but the status keeps reading 9-9-9-24-1 even if I set the parameters manually. I read somewhere else that Intel systems automatically go to a Command Rate of 1.

Over on the Gigabyte motherboard forum the only response I got suggested it might be that my board does not service PCI-e 3 cards well through PCI-e 2. Has anyone heard that being an issue at times? I get to PCI-e 3 if I upgrade my processor, but the i7-2700k already has a lot of horsepower so I kind of feel that an upgrade would be a bit of a waste.

 

[MTDEW]

The N = Command Rate, so its 2T.
Not sure where you read that Intel systems automatically go to 1T.
Both my 2500k and my 2600k systems are running at 2T with Gskill Sniper ram.

If indeed the Gigabyte board has issues with PCI-E 3.0 cards running in PCI-E 2.0 , that would certainly be a mobo/bios specific issue.
Since MANY people are running PCI-E 3.0 cards in PCI-E 2.0 slots.
Myself included, i'm running two 7950's in x-fire @ PCI-E 2.0 x 8x on my Asus P8z68-V Pro mobo.
And a GTX 670 on my Asrock z68 Extreme 4 mobo @ PCI-E 2.0 x16.
So If that was the case, i'd ditch the Gigabyte board first.

But that would be a very odd issue to have, so if i were you, i'd look into that more b4 drawing any conclusions.
And be sure you have the correct chipset and all other mobo drivers installed.

Did ya run prime 95 at all?
It may show errors with hardware other than the GPU alot quicker then running memtest.
Run some prime 95 Blend tests.

Use CPUZ and GPUZ to check your memory timings in windows and what your GPU is running at. (ie: PCI-E 2.0 @ 16x)

Also, testing memory is very time consuming.
Testing each stick individually is one thing.
If they all pass , you still are gonna have to test with all the ram installed together,since sometimes mobos have issues with all ram slots filled and the memory timings need to be lowered to prevent errors.
Although neither of mine do and both my systems have all slots filled with 16gb also.
This may not really be much of an issue any more since the memory controller is on the cpu now, but its something i still test when i fill all slots just to be sure.

And i agree 100%, no need to upgrade a 2700k.

 

[carlboyer]

Thanks for the additional suggestions. I'll try them this weekend. One thing, the mention of the card versus the system freezing. One more that one occasion, the screen froze while a task was running in the background - uninstall or Windows update - and if I left the machine run for a bit (could see the activity likes blinking and hear the hard drive banging away) the tasks were shown as complete after reboot.

 

[monkeydelmagico]

some bits never play nice together. Maybe sell card to buddy who was impressed with it and get something else?

 

[carlboyer]

boboftw, I did tried disabling hardware acceleration for Flash and that seemed to eliminate freezes after watching Youtube videos, but other things like opening Outlook and running the Performance Index continued to cause freezes. So on to testing.

To everyone, the idea of multi-hour memory testing is a bit daunting. So for starters I tried the blended Prime 95 tests albeit for short durations. The system passed the short Prime95 testing so I decided to try 3dMark.

First I tried my card in my old system, a EVGA 680i SLI board, with a Corsair HX620 PSU, E8600 processor and 6GB of RAM. The card ran under all conditions including 3dMark testing. So on to 3dMark in my rig.

When I first tried 3dMark, I found that I could not watch Hardware Monitor and 3dMark at the same time on one monitor as 3dMark wants to hold focus. So, I put my normal monitor (Dell 1908FP) on a DVI to HDMI adapter and connected it the HDMI port. Then I connected a Sony SDM-S71R to the DVI port using a VGA to DVI adapter. I ran 3dMark on the Dell and had Hardware Monitor on the Sony. The 3dMark tests ran fine in my rig. Curious.

So, with both monitors running, I decided to run all my other scenarios that caused freezing including Youtube videos in Flash player with hardware acceleration enabled. Lo and behold, no freezes! I then tried disconnection and testing with each monitor disconnected and wouldn't you know, with either one disconnected the freezes return! Plug the both back in and good bye freezes. What the...?

Thoughts?

 

[monkeydelmagico]

a cable problem? Seems like the only variable that changed.

 

[KingFatty]

What about a simple attempt - just increase your memory voltage? That might help if the memory is borderline unstable, and may compensate for any issues of the motherboard automatically changing around the speed settings.

 

[railven]

Is this system relatively new? Here is an experience I had with my system when I had finally upgraded to an SSD:

Everything was set as before, only difference was an SSD (system is in sig, at the time it was just 1x7970 and i5 2500k but mostly still the same).

Anyways, randomly I'd just get a light blue screen - my sound would still play, but I basically couldn't use the computer, had to power it off.

Long story short, a firmware upgrade for my SSD fixed the issues. I haven't had that issue ever again.

Just an idea.

 

[carlboyer]

Don't think it is just a cable issue as it fails with either monitor disconnected even if I leave my primary monitor on HDMI. Only with both monitors connected does it remain stable.

Trying the voltages for the card was my next step. I'll let you know how that goes.

 

[yacoub]

Does your friend's rig have a dual monitor setup?

My initial thought was heat issues, knowing that most Sapphire models don't have heatsink coverage on the voltage regulators. Perhaps your case is hotter than your friend's, and in the case of the single vs dual monitor variance, maybe the card runs just a little hotter in one scenario than the other and it's enough to introduce some instability?