Geekbench 6 - 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.
Here's a comparison to a completely untuned, stock 7950X with a Noctua U12A cooler and DDR5-6000 CL36:20C 265K vs. 16C 7950X with tuned 6200CL36 (SMT off): https://browser.geekbench.com/v6/cpu/compare/8495170?baseline=20758
20C 265K vs. Det0x's 16C 7950X with tuned DDR5-6600 (SMT off): https://browser.geekbench.com/v6/cpu/compare/8495170?baseline=17634
20C 265K vs. Det0x's 16C/32T 7950X with tuned DDR5-6600: https://browser.geekbench.com/v6/cpu/compare/8495170?baseline=20301
20C 265K vs. 13900KS 7600 MT/s: https://browser.geekbench.com/v6/cpu/compare/8495170?baseline=545
20C 265K vs. 13900KF 7600C34 MT/s: https://browser.geekbench.com/v6/cpu/compare/8495170?baseline=14764
20C 265K vs. 12900KS 6000 MT/s: https://browser.geekbench.com/v6/cpu/compare/8495170?baseline=2861
Text processing seems to be the forte of Arrow Lake (good for compilation workloads I guess?).
Even against 32 threads, 265K scores a win in text processing subtest.
265K totally owns the 12900KS so looks like a lot of old Intel users can look forward to a decent upgrade in performance.
Stock 7950X is great at file compression and compilation but can't touch 265K in text processing. Wonder what's contributing to Arrow Lake's text processing prowess. Probably the Skymonts.Here's a comparison to a completely untuned, stock 7950X with a Noctua U12A cooler and DDR5-6000 CL36:
overall 9950x looks to be faster in almost everything, and totals are definitely 9950x. And no avx-512 in there ???I like that 265K is giving almost similar performance to 9950X in HTML5 Browser, Text Processing, Object Detection, Background Blur, Photo Filter and Structure from Motion MT subtests at a much lower price point.
This is why I love competition.
Why would that be Skymont given that it also leads in ST?Stock 7950X is great at file compression and compilation but can't touch 265K in text processing. Wonder what's contributing to Arrow Lake's text processing prowess. Probably the Skymonts.
Good point. Need a P core only run to see how it fares then.Why would that be Skymont given that it also leads in ST?
We would need to see the code for text processing test to be able to tell, ideally with profiler running during the testStock 7950X is great at file compression and compilation but can't touch 265K in text processing. Wonder what's contributing to Arrow Lake's text processing prowess. Probably the Skymonts.
vs. 9700X: https://browser.geekbench.com/v6/cpu/compare/8495170?baseline=8488442overall 9950x looks to be faster in almost everything, and totals are definitely 9950x. And no avx-512 in there ???
Since the GB6 author claims that these are real world representative tests, I think we should take his word for it and just test the 265K in some real world text processing task coded in any language of choice and see how it does against Ryzen. No need to see GB6 codeWe would need to see the code for text processing test to be able to tell, ideally with profiler running during the test
That's an overclocked 9700X. Not a good comparison.Great showing by the 9700X in ST but the 265K turns in some good MT wins against it
I chose it on purpose to show that an above average 9700X is having trouble in some MT tests against 265K. Quite a bit of trouble. AMD made a booboo here. They needed a 10-core single CCD this gen but decided to be cheap again. This is the price they paid. Not total domination over Arrow Lake. And like I've been saying, some people would prefer 20 physical cores rather than 8C16T due to SMT not lending itself well to some workloads.That's an overclocked 9700X. Not a good comparison.
Increasing the amount of cores would not make much of a difference in GB6's "MT" anyway though.I chose it on purpose to show that an above average 9700X is having trouble in some MT tests against 265K. Quite a bit of trouble. AMD made a booboo here. They needed a 10-core single CCD this gen but decided to be cheap again. This is the price they paid. Not total domination over Arrow Lake. And like I've been saying, some people would prefer 20 physical cores rather than 8C16T due to SMT not lending itself well to some workloads.
Isn't the 265K (250W; $404 at Newegg) closer in price to the 9900X (162W; $429 at Newegg) than the 9700X (88W; $325 at Newegg)? Can you find that comparison?I chose it on purpose to show that an above average 9700X is having trouble in some MT tests against 265K. Quite a bit of trouble. AMD made a booboo here. They needed a 10-core single CCD this gen but decided to be cheap again. This is the price they paid. Not total domination over Arrow Lake. And like I've been saying, some people would prefer 20 physical cores rather than 8C16T due to SMT not lending itself well to some workloads.
Then are you saying that Intel managed to make a better "real world" multithreaded CPU than AMD? The scores are right there. +30% in two subtests and approaching 40% in Clang MT. If more cores won't help AMD here, then did AMD lose the IPC race in these subtests?Increasing the amount of cores would not make much of a difference in GB6's "MT" anyway though.
Isn't the 265K (250W; $404 at Newegg) closer in price to the 9900X (162W; $429 at Newegg) than the 9700X (88W; $325 at Newegg)? Can you find that comparison?
Sweet indeed, I got my 9900X at the current 9700X price, lol. As an added bonus, it craps on Core Ultra 265K chips in Geekbench and gaming.
Verdict: Josh wins!
But I expect the 265K to see a swift price cut as it will not be preferred by gamers. So if it gets cut to $350 then it would look like the better value. But yes, right now, @Josh128 managed to secure a sweet deal
But we don't have any information about what it costs Apple for their SoCs. We don't know what the defect rate of their silicon is. Maybe Apple products are expensive due to low yields which is why Apple skimps on RAM and storage to try to make up for most of the cost that way. We've already seen how much trouble Qualcomm had trying to mimic Apple's design approach with the actual M1 architect. One can espouse the benefits of Apple's approach all they want but until we know more, Apple might be sacrificing more wafers than AMD to give their users record computing performance.Yeah that's not good for x86, and M5 is coming next year (presumably).
But we don't have any information about what it costs Apple for their SoCs. We don't know what the defect rate of their silicon is. Maybe Apple products are expensive due to low yields which is why Apple skimps on RAM and storage to try to make up for most of the cost that way. We've already seen how much trouble Qualcomm had trying to mimic Apple's design approach with the actual M1 architect. One can espouse the benefits of Apple's approach all they want but until we know more, Apple might be sacrificing more wafers than AMD to give their users record computing performance.
Yeah - too bad Apple doesn't offer overclocking possibilities. Actually I'm curious why nobody doesn't have tried to hassle with those Apple chips clock schemes - it's fastest cpu core out there afterall.Yeah that's not good for x86, and M5 is coming next year (presumably).
Also worth mentioning that 4.5GHz is 99% not the limit of P core, I'm sure at the very least it could clock at like 4.8GHz or something like that, even if double the power would be needed to do that. x86 players are using obscene amounts of power for that single core frequency, Apple could do the same, although obviously the core wouldn't reach 5.7GHz or similar frequencies.
Yeah - too bad Apple doesn't offer overclocking possibilities. Actually I'm curious why nobody doesn't have tried to hassle with those Apple chips clock schemes - it's fastest cpu core out there afterall.