Me and someone Ik have been stuck around the same score of 33K, guess it’s because we're on Nvidia?As a long time lurker I feel like I can contribute a little bit to this thread.
View attachment 109841
Maybe your 14900KF is thermal throttling.Me and someone Ik have been stuck around the same score of 33K, guess it’s because we're on Nvidia?
Guess AMD went conservative on the clocks for this SKU, and looking at the already leaked numbers for Arrow Lake, they will face zero pressure from Intel this generation.With 5.2 Ghz Max Boost, rather than 5.4-5.5 GHz some people expected, 7800x3d owners can sleep well, without temptations to upgrade.
Not reallyMe and someone Ik have been stuck around the same score of 33K, guess it’s because we're on Nvidia?
Almost perfect, as F1 cars go - they have to get to the finish line and then self-disassembleHad one run at 53.3k, but the clocks were pushed to high -> freeze at the outro-screen of all things 😅
With 5.2 Ghz Max Boost, rather than 5.4-5.5 GHz some people expected, 7800x3d owners can sleep well, without temptations to upgrade.
Is this 45k score at 5,13 GHz and 300W of power for 9950x default state of business with PBO on,or does it require further tinkering to get there (meddling with CO and other bios settings, in other words not guaranteed)?
Do I get this right, he made a show of overclocking Zen5 so it could match subambient cooled M4 in SPEC instead of simply enabling SIMD in the compiler options, where it would match it at stock (at least the FP part) ? [This assumes he used his standard compiler setup for SPEC].
Would make sense. I installed W11 24H2 and managed to hit 34K, but couldn't get any more out of it, even with RTX 4090. I'll try at some point with my venerable RX 6800, once I get it backMe and someone Ik have been stuck around the same score of 33K, guess it’s because we're on Nvidia?
Not only that, the GCC version used to compile SPEC in his testing, GCC 10.1, predates not even Zen 4, but Zen 3.Do I get this right, he made a show of overclocking Zen5 so it could match subambient cooled M4 in SPEC instead of simply enabling SIMD in the compiler options, where it would match it at stock (at least the FP part) ? [This assumes he used his standard compiler setup for SPEC].
Isn't he using Clang 14.0.6 but without specifying any arch so he gets x64 baseline for everyone? It's better than specifying march=native if he doesn't plan on updating the compiler.Not only that, the GCC version used to compile SPEC in his testing, GCC 10.1, predates not even Zen 4, but Zen 3.
lol
Looks like at least half of my "not a chance in hell" prediction came to fruition, the performance half. Now its completely up to AMD. "To milk or not to milk".Idk here ? xD "Not, a chance in hell" then 12h later xD when tuned 9800x3d is 25%+ faster vs max tuned wcckek's best amd-owner's can do. But that meme was funny thou xD
View attachment 110017
10:58Isn't he using Clang 14.0.6 but without specifying any arch so he gets x64 baseline for everyone? It's better than specifying march=native if he doesn't plan on updating the compiler.
I think the twitter user Raichu was using GCC 10.1, David Huang is using GCC 12.3 but he is trying to make sure his march options make sense and Chips&Chesse are using GCC 14.2 with --march,mtune=native which is the best option for any chip as long as it is recognized by the compiler but it will force them to update the compiler for each released chip that GCC 14.2 does not support [or manually enable missing features].
Oh, I haven't noticed. My bad. That is why I was saying in another thread the compiler configuration should be shared in visible place together with the results
He's been all over the place with his choice of compiler. He used Clang 14.0.6 for his LNL review as you pointed out. But his Raptor Lake reviews used GCC 12 IIRC.Oh, I haven't noticed. My bad. That is why I was saying in another thread the compiler configuration should be shared in visible place together with the results
Thanks for pointing this out.
I mean everyone that expected way more than 10% was obviously wrong. ZEN5 is 3% faster than ZEN4 with same RAM and clocks have a 0.5-0.7 scaling. So it's 3% from Arch and 7% from Clocks at best. With 5.2GHz boost we're looking at 3% from Arch and 3% from Clocks, so ~6%.
You're judging a leak which may not even be accurate or real information against.... another leak that may or may not even be accurate or real information.AMDs marketing has a proven track record for setting a range of expectations and having the product come in at the absolute lowest possible point on that range. They literally leaked 5.2GHz all core Cinebench R23 a couple weeks ago. Based on that, most logically thinking people were assuming 5.3 to 5.4 max boost for 9800X3D. AMD set these expectations with that leak. Coming in at only 5.2 is disappointing solely because of their own crappy marketing. All they had to do in the run up was tease a "modest max clock boost", or dont mention clocks at all, but thats not how they roll, lol.
Who said that was AMD? It was clearly an overclocked CPU which was most probably (to put it mildly, I'd even say definitely) a dual CCD part with one chiplet disabled (otherwise it wouldn't require hiding the CPU name etc) which has higher (or just different) fmax as compared to non-ES SKUs including 7800x3d. As Dom said, his pre-release ES "9950x" has different fmax and v/f curve as compared to the final productsAMDs marketing has a proven track record for setting a range of expectations and having the product come in at the absolute lowest possible point on that range
The 5.2GHz all core leak was most definitely done with AMDs permission. There are NDAs for things like this, and its extremely evident that it was a calculated / planned leak.Who said that was AMD? It was clearly an overclocked CPU which was most probably (to put it mildly, I'd even say definitely) a dual CCD part with one chiplet disabled (otherwise it wouldn't require hiding the CPU name etc) which has higher (or just different) fmax as compared to non-ES SKUs including 7800x3d. As Dom said, his pre-release ES "9950x" has different fmax and v/f curve as compared to the final products
You can also make 7800x3d and 7950x3d to run close to fmax on x3d chiplet, so what's different now and why we should be worried etc? This 5.2 leak was just to curb expectations if people see the 5.6 ghz benchmarks and assume that it'd be 100% possible to do for everyone.The 5.2GHz all core leak was most definitely done with AMDs permission. There are NDAs for things like this, and its extremely evident that it was a calculated / planned leak.
Nothing about this “official leak” (whatever the hell that means) looks reputable to me. The graphic style (font, layout, image) doesn’t match any other AMD material, the verbiage itself reads very casual/unprofessional and frankly grammatically incorrect. At worst, this is entirely fabricated, at best it might be sourced from some official channel, but in neither of those scenarios does this appear calculated/planned.The 5.2GHz all core leak was most definitely done with AMDs permission. There are NDAs for things like this, and its extremely evident that it was a calculated / planned leak.