Geekbench Blog
www.geekbench.com
Weird choice of baseline CPU and even weird is that the baseline score is 2500.
i7-12700 does hardly 2000 in GB5 with the fastest DDR5.
12 threads help AMD CPUs with all cores strong, while Intel needs to run 4 or even 6 of these threads on weaker E cores. In this particular thread number advantage of the AMD CPUs is substantial.... After another quick test where I upped my polling rate, I take back this statement somewhat. Most tests seem to use ~12 threads on my CPU with Clang using ~20
Lightly multithreaded loads favor AMD, because they have all cores large, fewer threads hit weaker E cores on Intel CPUs, but not that many of them to reap the full benefits of them.
And in spite of this 13600K at stock has 88% ST and 77% MT performance of the 7950X at half the price. Impressive value.
No problem, I am proud of myself of resisting to pick up the 13900KS I ordered and going instead with the rational solution... Although this benchmark forced me down to 5400 MHz, which seems sooo slooow compared to 6 GHz of the 13900KS. They still have it in the shop. Ready for sale.... Oh no, it is still tempting me.Hmm
13900K : https://browser.geekbench.com/v6/cpu/409
13900KS: https://browser.geekbench.com/v6/cpu/545
I can play that game too, this time comparing to intel's top of the line parts:
"And in spite of this 13600K at stock has 86% ST and 76% MT performance of the 13900K/13900KS at half the price. Impressive value."
Yeah. For me. And for another 7 999 900 000 people on this planet 128 threads are useless.128 threads is almost useless ?????
This new benchmark is apparently useless for comparing server CPUs... Is this a problem? I believe that servers can be specialized and somebody building these servers and wanting to compare the CPUs for them needs to run specific benchmarks which correspond to what the server is going to do.No Doubt MT scores have taken a huge hit.
3200MT/s with bad timings as it is ECC.What memory speed are you running? I'm running 3600 MT/s with somewhat optimized timings.
Not only Servers, but HEDT aswellThis new benchmark is apparently useless for comparing server CPUs... Is this a problem?
On one hand it makes GB6 useless for servers/workstations. But it does make it even better for normal consumers. Really these many core CPUs do only provide minimal gains for most people.Not only Servers, but HEDT aswell
No, it is dead, do not mention it. Forget about it. Think WORKSTATIONS now!Not only Servers, but **** as well
On one hand it makes GB6 useless for servers/workstations. But it does make it even better for normal consumers. Really these many core CPUs do only provide minimal gains for most people.
My scores with 13900KF at stock clocks with DDR5 7600 CL34:
My scores with 13900KF at stock clocks with DDR5 7600 CL34:
Kinda interesting to compare it with DDR4 4000C18 13900KS 8C8T, undervolted stock:
Coz heavy multithreaded would favor AMD!
WOW. Are they being paid off by Intel?
They changed the way multicore works to be more in line with how it actually works for desktop workloads.From the TR PRO comparison runs, its clearly not a good MT run benchmark. 128 threads is almost useless ?????
Kinda interesting to compare it with DDR4 4000C18 13900KS 8C8T, undervolted stock:
Micro-Star International Co., Ltd. MS-7D25 - Geekbench Browser
View attachment 76484
I think B-Die stuff would perform even better, this system needs 64GB unfortunately, was suprised it actually ran 4000.
View attachment 76485
Its useless for almost anything with more than 12 threads. Thats like 13900k, 7950x, ALL HEDT and workstation CPUs, and all servers. Its leaves just mid end desktop and below is all that its good for.This new benchmark is apparently useless for comparing server CPUs... Is this a problem? I believe that servers can be specialized and somebody building these servers and wanting to compare the CPUs for them needs to run specific benchmarks which correspond to what the server is going to do.
and my 2 256 thread systems ? I can't compare them. If what it does only works for middle of the line desktop workloads, what good is it ?They changed the way multicore works to be more in line with how it actually works for desktop workloads.
I wouldn’t say it doesn’t scale, I bet 16 < 32 < 64 cores if all other things are equal (uarch, clocks)
I may play with it later.
You should consider going for a DDR5 setup. 64gb of low latency DDR5 working at 6000-6200 in my AMD system, and I am sure Intel users have even faster stuff.