• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

New build has terrible graphical performance

spengos said:
Ok, I've been running the test for 10 mins... I've got 4 Worker Windows open, and I'm guessing these represent each CPU thread/core?

The first three Worker Windows have been finishing tests at pretty much the same time(within a couple seconds), however the last window Worker#4 is stuck on test 1. What does this mean? A dodgy core?
Click to expand...

I don't know if they're assigned to each core individually, but just to make sure you can set the core affinity via task manager by clicking on the task and choosing "set affinity".
 
kylew said:
I don't know if they're assigned to each core individually, but just to make sure you can set the core affinity via task manager by clicking on the task and choosing "set affinity".
Click to expand...

Yes, all CPU cores are selected under the Prime95 process.

It also specifies each core at the top of each worker window.

If that 4th core is dead then it all makes perfect sense. Explains the -25% cinebench result and probably explains why the CPU > GPU performance sucks.
 
It would most definitely explain the Cinebench results, but the CPU > GPU performance issue is still a bit confusing. How about setting 3D Mark 06's affinity to 2 cores that you think are fine, give it a run through and see how it goes? You'll know after one run if that's the issue.
 
kylew said:
It would most definitely explain the Cinebench results, but the CPU > GPU performance issue is still a bit confusing. How about setting 3D Mark 06's affinity to 2 cores that you think are fine, give it a run through and see how it goes? You'll know after one run if that's the issue.
Click to expand...

I found a setting in the BIOS where I can manually activate/disable each core. I disabled the 4th core and then ran the cinebench GPU test and its much improved. I also tried 3DMark 05 and 06 and they're a little better but still not what they should be. Also tried Dragon Age and that seemed much better too. Still not 100% but I guess you have to assume the missing core is having some impact.

In any case i'll be sending the CPU back to Overclockers for a replacement. Sent the GPU back last week and now the CPU is going back too. Ah well, at the very least I can now do something about it. Was getting quite worried that I was going to end selling it all for spares!

Many thanks guys. Especially bigjimmy.
 
Hey sorry not been keeping up with the thread been watching a film and tweaking my system.

So indeed it does look like a defective core before you send it away can you try get us some screenshots of prime 95 with it stuck on test 1 or falling behind and core temp in the same view same screen shot my also come in handy if sending the cpu back.

Also just run through bios and make sure there is no option for individual core boost control and make sure all settings appear normal. Just want you to be extra sure before sending it back but it does pretty much tie in with the lack of cpu performance in the benchmarks that you ran. Especially cinebench which was a canny 25% less performance.

Anyway please keep us informed on what happens as we all hate it when no one lets us know how they got on or if they got a refund/exchange etc etc
 
Run OCCT, it will more easily fail if there is an issue with one of the cores, rather than continuing like Prime95 does.
 
Speed said:
Run OCCT, it will more easily fail if there is an issue with one of the cores, rather than continuing like Prime95 does.
Click to expand...

Why go an extra step it's obvious the core is defective already both from the prime 95 results and the fact after shutting that core down the performance increased. I think the core is fully working but it's not proccessing the data fast enough so to speak you could test this theory by running prime 95 but waiting half hour or so to see if that core does indeed complete test 1

by the way you never mentioned what the core temps were spengos when running prime.
 
SPENGOS READ THIS IMPORTANT



Just had a thought take off the cooler and remove the cpu and check and make sure there is no fluff/gunk stuck in the mobo socket as it could theoretically stop a core from working properly also check for bent pins most common problem according to google is people getting balls of polystyrene stuck in the socket holes or thermal paste...........

Just thought i'd share before you send it back as it is a valid possibility

If you lived closer i'd let you try my 955 in the socket to make sure it's ok
 
bigjimmyauk said:
Why go an extra step it's obvious the core is defective already both from the prime 95 results and the fact after shutting that core down the performance increased. I think the core is fully working but it's not proccessing the data fast enough so to speak you could test this theory by running prime 95 but waiting half hour or so to see if that core does indeed complete test 1

by the way you never mentioned what the core temps were spengos when running prime.
Click to expand...

Nothing is ever certain or obvious when it comes to computers, its an hour test but likely won't get that far.


bigjimmyauk said:
SPENGOS READ THIS IMPORTANT



Just had a thought take off the cooler and remove the cpu and check and make sure there is no fluff/gunk stuck in the mobo socket as it could theoretically stop a core from working properly also check for bent pins most common problem according to google is people getting balls of polystyrene stuck in the socket holes or thermal paste...........

Just thought i'd share before you send it back as it is a valid possibility

If you lived closer i'd let you try my 955 in the socket to make sure it's ok
Click to expand...

Good idea, bent pins could cause this. When you remove the CPU, check the socket and the underside of the CPU for damage, or any dirt, thermal paste etc that could be causing problems with the CPU and its interaction with the board.
 
Last edited:
Speed said:
Nothing is ever certain or obvious when it comes to computers, its an hour test but likely won't get that far.




Good idea, bent pins could cause this. When you remove the CPU, check the socket and the underside of the CPU for damage, or any dirt, thermal paste etc that could be causing problems with the CPU and its interaction with the board.
Click to expand...

I'm actually now taking bets on something interfering with the socket lol
 
Pure guess here (because I haven't kept up with the AMD market of late), perhaps it is a tri core that has been unlocked rather than a true quad?

Given the result of your cpu it seems to be cpu related, and it could be something as simple as no heatsink contact on one part of the cpu.
 
Here's a screenshot of the Prime test running. This one didn't run for long as I just did it to take a screenshot. But you can clearly see the 4th core isn't running. The tests I ran last night went on for much longer and the 4th core didn't complete a test in any of them.

51945018.jpg



I did try that OCCT program too but it didn't find any errors. Not sure what's going on there. Why would in not work in Prime but be fine in OCCT?

I did some further testing with Cinebench GPU test, playing around with the affinity of the cores to see what happened.

All 4 cores on = Poor performance
Cores 1,2 and 3 on = Normal
Cores 1 and 2 on = Normal
Cores 1 and 4 on = Poor performance

Also, while that 4th core was enabled the GPU usage shown in GPU-Z was pretty much 0% the whole time. There were a few small spikes of usage but very low(5%) and a good few seconds apart. On all the tests where the 4th core is disabled I get a nice constant usage of around 50%, with no dips or spikes. So that 4th core is clearly interfering with the CPU > GPU.

I did also pull everything apart again this morning and checked the CPU socket and CPU pins. All seemed ok. Although, with all this messing about I'm lossing a little of the thermal paste between the CPU and heatsink. I'm pretty sure there's still good contact though as the two still stick together quite tightly and there's no visible gaps or holes in the layer of paste.

I've got a mate who I think might have an AM3 mb so I might try and get round to his at lunchtime to see if we can do some testing. If I can try my CPU on his board and vice versa then it should be fairly clear where the problem is.
 
yup definitely cpu or mobo then you thorough testing clealy shows this

good idea to try the cpu in another mobo occt would pass fine it's just the core is running slow. If it was dead it wouldn't even show on the test the fact it does show but is running poorly make me think it's just a messed up chip or mobo.

try swapping cpu to another mobo if the same thing happens send the cpu back and tell them to read this thread for detailed info if they need to. if it works fine in another mbo then send the mobo back instead
 
Certainly looks like you have the root of your issues.

Try to downclock the CPU to under 3ghz and leave the 4th core on and see if it suddenly works okay. I have heard of this kind of thing happening to people. Ultimately they RMA'ed the cpu because all cores were not functioning properly at stock speeds but downclocking would prove the 4th core is dodgy.

It is also possible OcUK managed to get a fake supplied to them not their fault but as someone said could be a Tricore with a bad core but unlocked.
 
M3N-HT Deluxe/Mempipe

3-Way SLI provides Scalable Performance
– AMD AM2+/AM2 Platform
– Support latest AMD Phenom ™ Quad-Core Processor
– Dual-Channel DDR2 -1066/800/667/533
– 3 x PCIe 2.0 Gfx with NVIDIA 3-Way SLI™ Support
– NVIDIA Hybrid SLI™ and DX10 Support
– 10°C Cooler! Revolutionary Memory Heat Pipe
– Instant Internet Access with Express Gate SSD
– 8+2 Phase Power Design
– Support 45nm Phenom™ II CPU
 
yep it does support your cpu but it will probably need the latest bios update. Try it and see but it's definately on the cpu support list so it should work fine.

let me know how you get on
 
Well the results are in. Looks like a problem with the motherboard.

Tried my CPU in his system and it worked fine. Tried his CPU in my system and the 4th core wasn't working properly, just like it did with my CPU.

We only had time to test it on Prime95 so I'm hoping that's a good enough test.

I've checked the socket on my mb with a magnifying glass and i can't see anything obstructing the pins.

Im going to start an RMA and get it sent back. I was hoping it would be the CPU that was faulty... Its a smaller box to send back. :p
 
You must log in or register to reply here.
Back
Top Bottom