Dolphin (emulator) - the best stability tester

[nyker96]

I think the question if dolphin itself is stable is a good one. However, if dolphin runs fine on your systems for hours and hours on say a lower cpu speed or higher voltage, then it crashes doing the same program within minutes on a faster cpu setting then I'd suspect the unstablility comes from the cpu.

 

[cmdrdredd]

About 84c IBT Max temps.

 

[coffeejunkee]

Of course I am. The temps vary a bit between runs but those are pretty much what I'm seeing.

IBT is of course higher

Intel Burn Test and LinX are the same, aren't they? I assumed they were since my temps are about the same in both.

Needs more explaining, because both Linx and IBT are indeed the same, just a gui for Linpack.

OT: I'm experiencing the same Yuriman, Linx and Prime can run for days but still get crashes when gaming. My weapon of choice for stability testing: Forged Alliance.

 

[cmdrdredd]

Needs more explaining, because both Linx and IBT are indeed the same, just a gui for Linpack.

OT: I'm experiencing the same Yuriman, Linx and Prime can run for days but still get crashes when gaming. My weapon of choice for stability testing: Forged Alliance.

Should be the same but when IBT is set to max or high load it heats up more than any setting I can find in LinX.

 

[coffeejunkee]

Well, that's strange if you ask me. Just ran Linx and IBT at 4.1GHz, both at all/maximum setting (same amount of ram, little over 7GB):

Linx: 71C, 114.8 GFlops
IBT: 69C, 113.0 GFlops

 

[Idontcare]

Well, that's strange if you ask me. Just ran Linx and IBT at 4.1GHz, both at all/maximum setting (same amount of ram, little over 7GB):

Linx: 71C, 114.8 GFlops
IBT: 69C, 113.0 GFlops

As with all things software, there are multiple revisions to both LinX and IBT.

And even if you grab two versions of each that are truly based on the exact same execution binaries, you can still get different results because the execution overhead for the GUI is going to be different between the two and you need to make sure they are running the exact same test # (same memory).

 

[cmdrdredd]

4096MB used in each test. Notice IBT max temp was 82c and LinX was 72c. I let each run about a minute to heat up. Odd isn't it?

 

[Idontcare]

If you are fine in p95/ibt, you would be fine in anything. Also LinX shouldn't take long to find stability. I would say less than an hr and you will find true stability.

You really can't make blanket statement assertions like that, they are patently untrue.

There are over 700 instructions in the ISA, each one has its own sensitivity to voltages and temperatures, as well as sequencing of use.

IBT/LinX/P95 test a very small instruction subset of entire ISA. At best you can only conclude that those handful of instructions are stable, but you have no idea if the other instructions are stable at your rig's voltage and peak temperatures.

Like the OP, I too have a program that is not stable at temps/volts for which my rig is otherwise stable with LinX. It is Gaussian, a computational chemistry program, but it costs a bit of money (~$1k for the windows license) and so I would never recommend people use it instead of a free program like LinX.

The only way anyone can claim their rig is truly stable at a given Vcc and peak operating temperature is for them to create a program that tested each and every instruction in the ISA multiple times while simultaneously holding that CPU right near TJmax. The programs exist but they are not free (nor are they purchasable as neither Intel nor AMD are inclined to divulge their validation IP).

So until some geeks create such a test suite, and free at that, we are left with using power-viruses like OCCT and IBT for seeing if our OC's are unstable...but we can't actually use them to prove/verify that our OC's are stable.

 

[fastamdman]

So we can use them to see if we are unstable, but not stable. Makes a lot of sense, good read.

 

[Yuriman]

I decided to test to see what the minimum voltage necessary to get stable at 4.4ghz was. At 1.184v (as reported by CPU-Z) I was Prime stable and IBT stable for 20 loops, but Dolphin kept crashing within minutes. After about an hour of trial and error I ended up adding another +24mv of offset before Dolphin stopped crashing in the first 5 minutes. I crossed my fingers and went to bed, Dolphin was still running when I woke up 8 hours later.

Final stable voltage at 4.4ghz: ~1.200v.

 

[Idontcare]

So we can use them to see if we are unstable, but not stable.

Excellent summary :thumbsup: Spot on!

 

[sm625]

It is nothing but pure hubris and ignorance to suggest that linx, prime95, and IBT are the best and/or the only valid means of stress testing. There are many many programs that will instantly crash on a highly overclocked system that is "prime95 and IBT stable". I've seen it myself on more than two systems I've overclocked > 30%. The programs that cause crashes do not even need to be placing a heavy load on the cpu. Jesus the E6600 conroe computer I am on right now is fully prime95 and IBT stable at 3.3GHz. But it is not a stable system at that speed. Simple programs caused crashes that made me reduce the OC to 3.2.

 

[StrangerGuy]

Emulators are among the ultimate of unstable coding hackjobs and you want to suggest it as a good stability test? Gimme a break.

 

[Yuriman]

So we can use them to see if we are unstable, but not stable. Makes a lot of sense, good read.

This is more accurate, poor wording on my part.

Dolphin is more sensitive to overclocks on my chip than Prime or IBT. It will fail when they will not. It's a good test for instability, but as you say, it doesn't test for stability. No program does.

 

[SlitheryDee]

I'm bumping this thread because it pertains somewhat to a problem I've been having.

I've been trying to run Wind Waker on Dolphin for the past couple of days, but I've found the stability to be horrible. It never occurred to me that my overclock might be causing it because my system is stable otherwise. The OP mentions that raising the voltage basically fixed his problems, but I'm afraid to go any higher on my voltage due to thermal constraints.

I was basically wondering if anyone else had success fixing stability problems in Dolphin by reducing their overclock. If so, did you have to run at stock speeds for absolute stability or was a reduced overclock sufficient? FWIW I'm running a 2500K at 4.2 ghz on an ASUS P8P67 motherboard.

 

[TidusZ]

Emulators aren't exactly the most stable programs... For stability testing you need something that you know will be stable if the system is stable at baseline.

 

[felang]

I'm bumping this thread because it pertains somewhat to a problem I've been having.

I've been trying to run Wind Waker on Dolphin for the past couple of days, but I've found the stability to be horrible. It never occurred to me that my overclock might be causing it because my system is stable otherwise. The OP mentions that raising the voltage basically fixed his problems, but I'm afraid to go any higher on my voltage due to thermal constraints.

I was basically wondering if anyone else had success fixing stability problems in Dolphin by reducing their overclock. If so, did you have to run at stock speeds for absolute stability or was a reduced overclock sufficient? FWIW I'm running a 2500K at 4.2 ghz on an ASUS P8P67 motherboard.

Just run at a lower multi to test stability, for example 3.5 Ghz instead of 4.5 Ghz with same volts.

 

[Piroko]

I was basically wondering if anyone else had success fixing stability problems in Dolphin by reducing their overclock. If so, did you have to run at stock speeds for absolute stability or was a reduced overclock sufficient? FWIW I'm running a 2500K at 4.2 ghz on an ASUS P8P67 motherboard.

Somewhat related (my system isn't oced, but undervolted), I had stability issues with some older games a few months back. In my case, bumping Voltage by the smallest increment (iirc .025 V) already fixed my issues.

You might want to try base clocks just to see if your crashes are related to your OC at all. If you can confirm it, then put it back to 4.0 Ghz/same voltage, that might be stable already.

 

[SlitheryDee]

I anyone wants to know, the problem was indeed with my overclock. I've been running Dolphin with zero crashes for 2 days now at 4ghz. I doubt I would use Dolphin as a overclock test, but the OP at least seems to be right about it's sensitivity to overclocking.