[SOLVED] i9-12900k Very Low Multicore Performance

[buges]

Hi

I've just build myself a new system after deciding to retire my trusty old i7-7700k build as my main machine, So I went with the following, because I'm not a gamer.

I9-12900k
Gigabyte Z690 UD ddr 4
850w psu
Some Crucial 16gb I had already
Just using internal gpu for the moment

installed Windows 11 and ran Geekbench 5 and was shocked at my multicore perfomance.

Geekbench puts my Single-Core Score at 1847 and Multi-Core Score at 9658

According to Gigabytes hardware monitor software the processor went from 4900ghz to 5100ghz maybe once or twice for a split second, processor drew max 145w
and max temp is 52c.

Here's want i have done so far. Reset bios and loaded optimized defaults, removed the battery and loaded optimized defaults

With a 850w psu and max temp of 52c I've ruled out lack of power and thermal throttling.

I've build at lot of systems over the years, but I'm pulling my hair out now, Here's some pic's from the bios if anybody spots anything.







Thank in advance

 

[mdd1963]

what is your default RAM speed, and, what have you selected within XMP?

52C max? (or is that 'above ambient'?) seems a tad low if running at 4.8-4.9 GHz under load)

Using HWmonitor (installed and opened for constant monitoring) and CPU-Z (bench/stress CPU) for inducing a load, run the latter for 6-10 minutes...; what clock speeds and temps are sustained by the P-cores and E-cores? (Be sure to hit 'stop' in CPU-z when done to terminate the load)

 

[buges]

Here's the screenshots from HWMonitor. CPU-Z and Gigabytes Hardware monitor. These were taken after 10mins of running CPU-Z's Stress.

See what you mean about 52c being low, looking at my cpu temp after 10mins, That MSI cooler is being returned, I might have enough spare parts just to knock up a better temp cooler, the reason I said 52c was because that's the highest it ever got during multiple Geekbench tests.
my ram is only 2400 no XMP and I only OC it to 2500. Only have one 16gb module installed I put the other one in from my old machine to try and put them in to dual channel mode, but my MB wouldn't recognize them in that mode, they are identical so don't understand why.

 

[mdd1963]

HWMonitor indicates much more realistic temps...(I suspect your reading of 54C was indeed perhaps the measured differential above ambient, or similar) and your sustained clock speeds look normal across the board.

Running only a single RAM stick in single channel, however, will indeed significantly drag down assorted scores in various benchmark comparisons, cutting memory bandwidth in half, and, lower speed RAM at that. (3200 MHz is the new normal, and many shoot for 3600 MHz..) I suspect that is your only real issue.

 

[buges]

Thanks for the reply

I'll buy some new ram, and post the results in a couple of days.

 

[TerryLaze]

Go here to the turbo power limits, it seems that it is "limited" to 145W for long time boost (intel guideline is 125W) anyway since you have power and cooling to spare you can increase that until you are happy.
You can use intel extreme tuning utility from windows to do that so you don't have to reboot all the time.

 

[uWebb429]

@buges
Set both the PL1 and PL2 turbo power limits to the maximum 4095W. This is how to get maximum performance out of these CPUs.

Run a reputable benchmark like Cinebench R23. Set it to do a single test, not a 10 minute test. Lots of 12900K users have posted their results online so you can compare your results.

Try using HWINFO instead of HWMonitor. Make sure your CPU can maintain full speed for the entire test.

 

[mdd1963]

It seems clear that the OP's single stick of RAM at slow speeds will cause a hit to his scores in almost any comparison, so, I'd see little (ok, zero) reason for any sustained cooling torture tests at this point....

(As the results already show 93C under CPU-Z within 10 min, it seems quite clear results of sustained Cinebench runs will likely cause throttling. Since their debut, 12900Ks are not really known for running at 4.9 GHz sustained at 53C . I suspect we will find not-entirely-optimal cooling solutions deployed in there somewhere, but, as a cooling solution currently deployed was not mentioned yet, I will withhold judgement, but, ....the 93C already indicates 'issues'...)

Running even 10 minutes wlll heat the CPU nicely, and allow us to judge not just the Intel-standard boost duration/clock speeds/TDP compliance behavior over the first 2-3 minutes, but, what will happen after a few minutes when/if TDP constraints would typically kick in on some mainboards.... (With the current cooling solution, those limits might come in handy)

 

[buges]

Sorry for the late reply.

Thanks everybody for the help, I bought some decent ram & now my speed seems to be about the same as every one else's on geekbench's website.