Question 9700K after overlock - same benchmark score

gipper53

Member
Apr 4, 2013
76
11
71
I just bought a new 9700K to upgrade my 8700K for S&G (price was good). My 8700K has been OC'd to 5.0GHz with Hyperthreading "off", as my main program (Photoshop) actually benchmarks faster with it off. I'm using a MSI Z370 SLI Plus mobo, rest of the specs are in my signature.

A basic test I did to compare (RAW to JPG export) took 1:12 with the OC'd 8700K. The stock 9700K with only XMP enabled did it in 1:03. So I'm thinking "Great, about 15% faster before overclocking". Right now I've got it at 5.0@1.33v, -2 AVX. Seems stable and temps have been good with basic stress testing. But when I re-ran the benchmark, I got exactly the same score. And this was a benchmark that previously showed good improvement with overclocking. It also maxes all cores out when running. So I'm a bit baffled why the OC didn't improve this here? Any ideas?

I did the overclocking following MSI's guide:
 

arandomguy

Senior member
Sep 3, 2013
556
183
116
Did you monitor clockspeeds? XMP on some motherboards also turns on MEC which would mean your 9700k would have already been boosting to 4.9ghz all core.
 

gipper53

Member
Apr 4, 2013
76
11
71
I was watching in RealTemp and they seemed to hover right around 4.6 I believe at stock. They were definitely clocking higher in RealTeamp after the OC.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
It also maxes all cores out when running.
Are you sure about that? Photoshop is littered with ST heavy workload fragments, one of the reasons why you discovered your 8700K scored better with HT off. You can test this easily by running one test with all cores limited to 4.8Ghz (and no AVX offset) versus 5GHz with -2AVX. Scores should be equal for a true MT workload.

My only suggestion would be to make sure you're not power limited. Check all power/current limits in the MSI BIOS (although MSI usually sets them very loose on their good motherboards), and also use HWINfo during a benchmark run to check for power limit flags (in the Sensor Tab).

Having said that, even assuming your test is truly well & uniformly multithreaded, and assuming the 9700K will hit that 4.8Ghz OC vs. 4.6Ghz stock, your OC score should be 4% better... meaning 1:01 - 1:02. You're trying to shave off 1 second.
 

gipper53

Member
Apr 4, 2013
76
11
71
@coercitiv

The export loads all cores up, at least according to task manager. It's one of the few operations in Photoshop that actually uses multi-cores decently. Not sure if there's a point it caps out, but so far it's seen good use on 6/8 core chips. I do wonder if it uses AVX instructions here because this operation did cap out at 4.8Ghz during the test (just ran it again). So that does explain why the same results.

But only a 15% improvement in scaling going from basically 6 to 8 of the same cores.
 

Attachments

  • Capture.JPG
    Capture.JPG
    482.1 KB · Views: 6