[resolved]Game/System crashes due to SLI or PSU?

[HokieRif]

Here's a softball for the forum...

 

I recently installed a second 7800gt in my system and everything appeared to be in good shape since the install - Windows XP runs fine, Aquamark and 3DMark ran beautifully, and I played around with a couple games (Half Life 2, C&C Generals, nothing too taxing). Last night the fun begins when I finally install CoD 2 and I crank all of the settings up to the max and enable the "SLI" feature in CoD2. While playing straight for 5-10 minutes I started to get some warnings from MBM5 saying my GPU and GPU ambient were beyond the threshold (it was set at 70c) - I ignore it and keep going... a few minutes later the whole system shuts down. A few seconds later I power the system back up, it gets to the BIOS screen, and it powers right back down. So, thinking I may have a heating issue, I wait some more, power it back up and it gets all the way to the Windows startup sequence before it crashses. I wait another 5 minutes, try again, and I'm finally back to Windows.

 

Long story short, the above sequence continues to happen and I suspect that I'm either running in to a cooling problem with my PSU (I kept the case open during my run earlier this morning and it managed to go a few minutes longer ~15 min.), or the addition of the new card has pushed the PSU to it's power limits and I need a few more watts to make this setup happen... So, my question to the group is am I thinking along the right path, and if so, do I place a house fan next to the psu to help with the cooling or do I need to look at a new PSU based on the equipment in my sig?

[tasr]

What are your temps? More important if you're air cooling your rig it will not get any cooler than the room ambient temp. Expect the CPU temp to be 10ºC or more above the room ambient temp.

 

I set my rig up for maximum cooling. If your HSF and chipset fan are plugged into the motherboard you can try the following.

 

Go into the BIOS, under PC Health Status and set the: “CPUfan Fully On…” > 26ºC and “NB Fan Fully On…” > 40ºC I also have the Shutdown Temperature – Enabled at 60ºC.

 

Basically this will set the fans to full speed (RPM) but will stop the fan throttling. I notice that when I do this my whole system will run cooler. I prefer the fan spinning faster than the throttling. If the throttling or noise is to loud adjust the temp up a bit.

 

Replacing the chipset fan with an Evercool VC-RE is going to help. A lot of people get a 5ºC - 10ºC drop in temp.

 

More case fan’s and high volume.

 

1. The key is your room’s ambient temperatures, the cooler the better.

2. Then you also need proper case air flow.

3. Keep the HSF and fan’s clean of dust and lint for efficiency. If you notice your temps rising and dust accumulation it’s time to clean.

 

 

At looking at you rig in your signature it draws a lot of power the PSU maybe an issue too. Your dual core SLI configured and probably overclocked too. I would have a minimal of [email protected] A more powerful PSU will not work so hard a put out less heat. It’s something to think about.

 

 

º¿º

[EllisD]

I think the problem could be with yout PSU, Antec's are not recommended around these parts. You have 2 x 19A 12v+ rails and the recommended for 2 x 7800GTX's is 34A. If you know anyone with a PSU that is on the recommended list, i would give that a shot. If not, i would look into:

OCZ GameXStream OCZ600GXSSLI ATX12V 600W Power Supply or

OCZ GameXStream OCZ700GXSSLI ATX12V 700W Power Supply. They have been the main recommendation since they have come out.

[HokieRif]

Well, that's 2/2 for people saying the PSU might be the issue (I like the OCZ suggestion - that would go well with my neon blue lighting theme in the case). I ran some tests last night (keyword here is "night) and increased the logging on MBM5 to get some more accurate heat readings during gameplay. I ran the game without issue for 5 mintue sessions and then read the logs coming out -

 

the peak temps I reached during gameplay are as follows:

CPU: 44c

PWMIC: 51

Chipset: 51

GPU Core: 77

Still no idea what the PSU temp is hitting, but it did feel like it was running cooler.

 

I also adjusted some of the performance settings in the NVIDIA configuration panel and ENABLED SLI Antialiasing... I think the CPU temp dropped 2c and hte GPU Core went up 1c when I did that...

 

Is that 77c on the GPU core a safe temp? I recall seeing somewhere that below 90c was safe on these cards, but someone correct me if I'm wrong. Other than that, I'm going to try pulling the lowe PCI card below the 2nd 7800 to see if that helps (reducing power draw/more clearance for air to pass over the 7800), and then I'll report back.

[HokieRif]

Just crashed the system again... here were the vitals from MBM5 seconds before the computer shut itself off

 

CPU 76° C

PWMIC 61° C

Chipset 61° C

GPU Core 79°

 

So which temps shouldn't I be worried about? The GPU core peaked at 81c minutes before the crash, so I doubt that's the problem. Looking back through the logs everything seemed normal up until 90 seconds before the shutdown - the CPU temps started to go up steadily from one minute up until the crash when the CPU temps went up 16c (from 50c to 76c).

 

So what in the heck could cause a spike like that over such a short period of time when the CPU was already running under decent load at 50c consistently for 10-15 minutes?

[EllisD]

76C on the CPU !?!?! Uh Oh.

 

What about fan speeds during this 90 second period? Are they still spinning? I could see the temp spike if they werent, but if you have a reasonable amount of case air flow, i doubt it would cause a 15C jump in 90 seconds.

[ebdoradz]

81°C for a GPU is normal, that what i get when gaming on my 6600GT, but 76 on a cpu is not the kind of thing you wanna see,

 

since you cpu is a stock speed, i would check if the heatsink is sit properly, using stock heatsink?? maybe replace the original thermal paste by some Arctic silver 5

[davidk21770]

76C on the CPU is WAY high. Probably what caused the crash. I'm pretty sure that the max rated temp is 57C for that chip.

I have my MB BIOS auto shutdown set for 60C.

 

As others asked...

What are the fan speeds leading up to this and did the CPU just spike or gradually rise?

 

EDIT: I also agree that the PSU is light weight for what you're running.

[SeanDude05]

AMD chips are only rated for safety up to 70c. What kind of cooling do you have one your CPU Hokie? Stock heatsink? Have you tried cleaning the cpu and reapplying a fine layer of AS5 or other thermal paste? Cleaned out the heatsink fins of dust?

 

I'm sure both those 7800s are fine, they can go all the way up to 90-100c. Have you tried to recreate the crashing with one video card? Is it only in SLI mode? It's starting to look like you need some serious case cooling.

[HokieRif]

AMD chips are only rated for safety up to 70c. What kind of cooling do you have one your CPU Hokie? Stock heatsink? Have you tried cleaning the cpu and reapplying a fine layer of AS5 or other thermal paste? Cleaned out the heatsink fins of dust?

 

I'm sure both those 7800s are fine, they can go all the way up to 90-100c. Have you tried to recreate the crashing with one video card? Is it only in SLI mode? It's starting to look like you need some serious case cooling.

 

Yup, the 7800s are fine (thanks again for card #2) - CPU is being cooled by a Thermalright XP-90c with a 90mm fan seated on top of the heatsink (running at 2000rpm constant). I'm also using AS5 for the CPU/heatsink (chipset has a nice layer on there as well).

 

and just as I'm writing this I have 2 instances of Prime running and the temp on the CPU has rawled up to 51c... then some major spikes that sent it all the way up to 60c... when I finally stopped both instances and began to scratch my head.

 

ugh... I've already got some good suggestions above for a new PSU (OCZ GameXtreme 700 is on the way), but adding a faster fan to replace the CPU fan I already have there isn't enticing (too much noise)... Anyone have a suggestion for some water cooling (CPU, Chipset, and 2 GPU)?

[davidk21770]

The last time that I had CPU temp spike like that, it was actually that my memory was giving me problems and needed a volt boost.

[wevsspot]

Since the A64 memory controller is on the cpu die, both your memory and your cpu really need to work in perfect harmony (including adequate voltage for both).

 

I'd check your BIOS and make sure that your putting at least 2.75 - 2.80v to your RAM, while you're there take a look and see what your BIOS reports your CPU voltage at, I'm guessing it's probably somewhere down around 1.36 - 1.38v. It might be worth a little bump in that department also.

 

Your CPU temp. under load is completely out of control. With decent air cooling you should be able to keep that down below 50C regardless of the circumstances. Your chipset, gpu and pwmic temps are about normal for an SLi board. It's time to pull that heatsink/fan and either lap, use AS5 and reseat, or replace all together.

 

Both the Artic Cooling A64 Pro or the TT BigTyphoon are good cooling choices if you want to stay with air. And both are fairly quite.

 

I wouldn't do anymore tasking work on your p/c until you get this temp. situation with your processor straightened out though. You're playing with a pretty pricey piece of cpu there.

 

Last but not least, the OCZ GameXStream 700W psu would be an excellent choice for your current build.