Results 1 to 2 of 2
Thread: How many benchmark runs
-
04-11-2015 01:51 AM #1
How many benchmark runs
Would you consider a setting to be stable?
SS
-
04-12-2015 01:18 PM #2
jab383 PC Specs Motherboard 24/7 rig : Maximus VI Extreme Processor i7 4790K Memory (part number) 16GB Mushkin Redline 2400 10-12-12-28 + 16GB Corsair Vengeance 2400 10-12-12-31 Graphics Card #1 AMD Firepro W5000 Sound Card M6E Supreme FX Monitor Dell U2413 Storage #1 Kingston SH103S3240G SSD Storage #2 Seagate ST1000DM003 1TB CPU Cooler Custom water loop, Delidded, Liquid Metal TIM Case CoolerMaster HAF XM Power Supply Corsair HX-750 Keyboard Logitech G710+ Mouse Logitech M705 OS Windows 7 64 Pro
- Join Date
- Feb 2014
- Reputation
- 107
- Posts
- 855
That's like asking: Define 'Stable'
If the benchmark passes one run, I would consider the setting to be "Realbench stable" since I can get a Realbench score with that setting.
Likewise if the setting passes Intel XTU - primarily Prime95 benchmark - then the setting is "XTU stable"
If the setting passes OCCT large data set testing for 3 minutes, I consider it "OCCT Stable"
The process continues for other benchmarks and the setting can be more aggressive when the test is less stringent.
In my experience, OCCT is the most stringent test and stresses the CPU moderately. It checks for data errors in either CPU or memory that occur in marginal settings. My 24/7 setting will always be OCCT stable and that's as close to unconditionally stable as I can get.
Realbench is second as a stringent test and is very gentle on CPUs. It can be used reliably with air cooling. A setting that passes the Realbench benchmark a few times - 3 to 5 - is probably 24/7 stable.
Prime95 is all stress and no test. It heats the CPU excessively and does not check for errors that show instability in marginal settings.
Aida64 is great for a monitor program and I use it exclusively. As a stability test ... not so much.
That's more than you asked, I know, but it's the best answer I can give.
Jeff