Overclock score worse than stock score?!

[TickleMyElmo]

So I put together a new build a couple days ago because my Asus p8p67 Deluxe motherboard went out. I purchased the 7700k, Asus Maximus IX Hero, and 16gb of Corsair Vengeance RGB 3000 ram. I overclocked the 7700k to 4.8 (1.28 vcore), the Asus ROG Strix 1080ti oc (+70 boost clock, +150 memory), and used XMP to use my ram at the stock speed of 3000. I ran prime95 26.6 small fft's for 20 min., no problems, with max temp. of 72c. Then I ran Asus Realbench stress test for 15 min., with the max temp of 75c during Handbrake, again no problems. Weird thing is, my overclocked cpu, ram, and gpu got a lower score than when it was stock. Why and how to fix this?

 

[Nebulous]

lol, 2 different programs will result in 2 very different scores. I'll bet if you ran those 2 benchies again you won't get the same score twice. Run some games and enjoy your new rig. Don't worry about what 2 different scores mean.

 

[trents]

So I put together a new build a couple days ago because my Asus p8p67 Deluxe motherboard went out. I purchased the 7700k, Asus Maximus IX Hero, and 16gb of Corsair Vengeance RGB 3000 ram. I overclocked the 7700k to 4.8 (1.28 vcore), the Asus ROG Strix 1080ti oc (+70 boost clock, +150 memory), and used XMP to use my ram at the stock speed of 3000. I ran prime95 26.6 small fft's for 20 min., no problems, with max temp. of 72c. Then I ran Asus Realbench stress test for 15 min., with the max temp of 75c during Handbrake, again no problems. Weird thing is, my overclocked cpu, ram, and gpu got a lower score than when it was stock. Why and how to fix this?

Neb, I think OP means he got lower scores on the same software after he overclocked than he did at stock frequencies.

Elmo, what about Cinebench scores when overclocked v. stock? And neither Prime95 or the Realbench stress test are benchmark programs. Realbench does have a benchmark mode, however, so use that instead. I would also run 3DMark at stock and and overclocked. Use some real benchmark tools.

 

[Alaric]

I've gotten lower scores with higher clocks before, usually due to instability somewhere. Try bumping the voltage to 1.3 on the vcore and test it again.

 

[TickleMyElmo]

I was just mentioning the p95 and Realbench stress tests were successful, atleast for the short time I ran them. I ran Valley and scores were up with the OC. Is 3DMark free?

 

[EarthDog]

Its free..limited function..but free.

Its likely instability somewhere...

 

[trents]

When you run stress tests, have CPU-z open to the CPU tab and watch the CPU multiplier. If it fluctuates under full load say in Prime95 then you may be getting throttling which would decrease your temps. You need to run Prime95 for at least two hours to confirm stability in your overclock. If your Valley scores were higher in overclocked mode than in stock mode it suggests the throttling is with the CPU rather than the GPU since Valley should really work the GPU but doesn't use the CPU much.

 

[TickleMyElmo]

Why would it throttle at only a max temp of 75c?
Running prime95 now, looking at cpu-z the multiplier and core speed drop to 4.6 every 5 or so seconds only lasting a 1/2 a second or less. Anything else I should be looking at?

 

[trents]

What is your AVX offset ratio set to in bios? That feature will automatically downclock the CPU when it encounters AVX/AVXII instructions sets which are found in Prime95 to maintain stability since these instructions sets are very demanding.

No, it shouldn't throttle at a max of 75c. My point earlier was that if it was encountering higher temps that the 75c might have been the result of the throttling.

Have you tried the Cinebench CPU benchmark at stock and overclocked? That one uses only the CPU so any gains by overclocking the GPU would not figure in.

 

[TickleMyElmo]

I set it to -2 following an overclock guide. Bought 3dmark, and Batman Arkham City both were $4.99. I will run it shortly. Also I will run cinebench.

 

[wingman99]

I can't wait to here the overclocking benchmark results. -2 down clocks the CPU 200Mhz.

 

[Nebulous]

Neb, I think OP means he got lower scores on the same software after he overclocked than he did at stock frequencies.

Elmo, what about Cinebench scores when overclocked v. stock? And neither Prime95 or the Realbench stress test are benchmark programs. Realbench does have a benchmark mode, however, so use that instead. I would also run 3DMark at stock and and overclocked. Use some real benchmark tools.

Ahh. I was saying that when I run benchmarks I never get the same score twice with the same benchmarks.

 

[Alaric]

Ahh. I was saying that when I run benchmarks I never get the same score twice with the same benchmarks.

Me either. Ever. LOL I go by "feel" interpreting the results. A dramatic or consistent change within a a range. Like a dynamometer, I'm looking for trends as opposed to hard numbers.

 

[TickleMyElmo]

I recieved a score of 9341 in Time Spy, not sure if that's good or not. Have you guys noticed in the Time Spy BM, the Large Hadron Collider, and aliens/demons. I personally don't believe in aliens, but I do however believe in demons. I think this benchmark might be telling us that demons will come from the LHC in the future (Time Spy).

 

[trents]

Back to the point, Is the Time Spy score significantly different in overclocked mode v. stock?

 

[TickleMyElmo]

Haven't ran it in stock yet, doing so now. Oh, and Fire Strike score was 21,487. Be back after setting to stock and redoing test.
What do you guys get in those tests?

 

[TickleMyElmo]

Ok, so the results are in. Time Spy - Stock/score 9125/graphics 10459/cpu 5297. Time Spy - OC/score 9341/graphics 10461/cpu 5814.

Fire Strike - Stock/score 20888/14125/28975. Fire Strike - OC/score 21487/15085/28916.

All tests were done with gpu set at +70 boost clock and +150 memory. Just the cpu and ram were changed from stock to overclock/XMP.

 

[Alaric]

Ok, so the results are in. Time Spy - Stock/score 9125/graphics 10459/cpu 5297. Time Spy - OC/score 9341/graphics 10461/cpu 5814.

Fire Strike - Stock/score 20888/14125/28975. Fire Strike - OC/score 21487/15085/28916.

All tests were done with gpu set at +70 boost clock and +150 memory. Just the cpu and ram were changed from stock to overclock/XMP.

Those would be the trends I was talking about.

 

[TickleMyElmo]

So all looks well to you then? No throttling or issues? doesn't seem like a big jump. Guess I expected too much. Gonna post Cinebench scores momentarily.

 

[TickleMyElmo]

Here are the Cinebench results. Cinebench cpu stock - 951 cb/single core cpu 195 cb. Cinebench cpu overclock - 1004 cb/single core cpu 208 cb. Gonna try and hit 5.0 with the cpu, we'll see what happens.