The Ryzen 9 7950X geekbench-tested beats Intel Core i9-12900K.

#6046934

D1stRU3T0R

2022-08-31 11:11

Slower than Intel 13 gen? Ahmm... interesting?

#6046953

Kaarme

2022-08-31 12:49

D1stRU3T0R:

Slower than Intel 13 gen? Ahmm... interesting?

Intel needs 24 cores to beat AMD's 16 cores in this chart. So, yeah, it looks pretty interesting.

#6046963

Aekold

2022-08-31 13:51

Kaarme:

Intel needs 24 cores to beat AMD's 16 cores in this chart. So, yeah, it looks pretty interesting.

Though with P/E cores, the thread count comes out to be the same. That said, AMD seems to be rocking single core performance! We'll have to see what the real power draw numbers look like. The 12900k was crazy and I don't see the next gen getting much better.

#6046989

nizzen

2022-08-31 15:47

Took a test with 12900k and DDr5 I don't know how it is possible to have such low 12900k score that this "leak" has..... This result is before starting to turn up the overclock 😛

#6046997

Horus-Anhur

2022-08-31 16:16

nizzen:

Took a test with 12900k and DDr5 I don't know how it is possible to have such low 12900k score that this "leak" has..... This result is before starting to turn up the overclock 😛

But that score is with your tweaked DDR5, right? Geekbench loves fast ram.

#6047005

Kaarme

2022-08-31 16:48

Pale:

It doesn't actually matter, how many cores Intel needs to beat AMD in a threaded workload. They just have a different architecture. What matters is that they do it within a comparable chip in terms of cost/die size and performance/watt. That 13900K also wins in single threaded performance equally impressive. Granted, the 3D cache chips will likely increase performance for AMD. And Intel will have to come up with a similar tech. But Intel appears to have the overall better idea about architecture with their big/little setup, with performance cores taking care of single-core workloads, while the efficiency cores ensure a lead in threaded performance as well. So I'm guessing AMD will have to Intel's way as far as overall architecture is concerned within a generation or two.

If the Intel CPUs were crazily power efficient fully taxed, thanks to the efficiency cores, I'd also say the architecture is a huge success, but in reality that doesn't seem to have been the case. The performance difference in the rumours/leaks so far also hasn't been truly significant. Obviously things need to move forward and change, so it's possible Intel's route is the best one forward. So far AMD's uniform cores seem to be competitive enough, though. Time will tell. It's going to be interesting to see the situation five years from now.

#6047019

Freeman

2022-08-31 17:22

Kaarme:

If the Intel CPUs were crazily power efficient fully taxed, thanks to the efficiency cores, I'd also say the architecture is a huge success, but in reality that doesn't seem to have been the case. The performance difference in the rumours/leaks so far also hasn't been truly significant. Obviously things need to move forward and change, so it's possible Intel's route is the best one forward. So far AMD's uniform cores seem to be competitive enough, though. Time will tell. It's going to be interesting to see the situation five years from now.

That thing can be 10Ghz and it would not matter if it was stuttering and causing restarts on every OS. General Hardware discussion is full of Ryzen. And will be. We will all see new posts of AM4 :"hey guys, what do I dooo? What's wrong with my PC?" and we will see AM5 with most likely new issues. Looking at how AMD approach release of new CPU it's to push sales. I dare you: Say one word about driver and software support! Ever heard of : " "Welcome to AMD's User Experience Program"? The one thing every user avoids? ... It will be interesting CPU and one to be careful to seriously look at.

Screenshot_2022-08-31-18-23-19-331_com.android.chrome.png

#6047032

Kelutrel

2022-08-31 18:10

Pale:

Intel appears to have the overall better idea about architecture with their big/little setup

No. Not even a bit imho. It would work if the E-cores were just slowed down versions, but they are also removing features from them. The OS scheduler will need tarot cards to predict if a certain thread can, or can't, run on an E-core and if there is any advantage in doing so. They are just trading performance for complexity on the OS atm. I would not put my bet on that big.little architecture in the long run, it is just a temporary solution while their process reaches a sufficiently small size to have 16 performance cores on a die without requiring 400W.

#6047041

nizzen

2022-08-31 18:38

Horus-Anhur:

But that score is with your tweaked DDR5, right? Geekbench loves fast ram.

Yes, and that's my point. They are testing more memory than the cpu. Looks like 12900k was using like 4800mhz xmp 😛

#6047048

schmidtbag

2022-08-31 19:07

Kelutrel:

No. Not even a bit imho. It would work if the E-cores were just slowed down versions, but they are also removing features from them. The OS scheduler will need tarot cards to predict if a certain thread can, or can't, run on an E-core and if there is any advantage in doing so. They are just trading performance for complexity on the OS atm. I would not put my bet on that big.little architecture in the long run, it is just a temporary solution while their process reaches a sufficiently small size to have 16 performance cores on a die without requiring 400W.

The features removed are things that the vast majority of software doesn't use or doesn't yield much benefit (HT). A single P-core is nearly 4x larger than an E-core, yet even with HT, it doesn't even offer double the performance clock-per-clock in most applications. Key word is "most", because while there are some where a P core makes a big difference, the E-cores are a better use of die space for the average desktop CPU. However, I do agree that this complicates the scheduler, but that seems to have been ironed out pretty well. Remember: logical threads cause scheduler complications too.

#6047050

Horus-Anhur

2022-08-31 19:10

nizzen:

Yes, and that's my point. They are testing more memory than the cpu. Looks like 12900k was using like 4800mhz xmp 😛

But the standard score for a 12900K un GB5 is around that value. https://browser.geekbench.com/processors/intel-core-i9-12900k

#6047060

winning.exe

2022-08-31 20:17

Kelutrel:

No. Not even a bit imho. It would work if the E-cores were just slowed down versions, but they are also removing features from them. The OS scheduler will need tarot cards to predict if a certain thread can, or can't, run on an E-core and if there is any advantage in doing so. They are just trading performance for complexity on the OS atm. I would not put my bet on that big.little architecture in the long run, it is just a temporary solution while their process reaches a sufficiently small size to have 16 performance cores on a die without requiring 400W.

E-Cores are not a temporary solution, you can very easily glean this from Intel’s roadmaps. In fact they will be releasing all E-core processors by 2024 😎. Moreover, AMD is developing their own E-core (first for the enterprise space) which will be released as Zen 4c (alongside Zen 4); whether or not these cores will come to consumer parts remains to be seen. Unless AMD wants to be completely humiliated in multi-core benchmarks from now until the end of time, they’re going to need E-cores in the consumer space 😛.

#6047064

JamesSneed

2022-08-31 20:27

Pale:

It doesn't actually matter, how many cores Intel needs to beat AMD in a threaded workload. They just have a different architecture. What matters is that they do it within a comparable chip in terms of cost/die size and performance/watt. That 13900K also wins in single threaded performance equally impressive. Granted, the 3D cache chips will likely increase performance for AMD. And Intel will have to come up with a similar tech. But Intel appears to have the overall better idea about architecture with their big/little setup, with performance cores taking care of single-core workloads, while the efficiency cores ensure a lead in threaded performance as well. So I'm guessing AMD will have to Intel's way as far as overall architecture is concerned within a generation or two.

What matters is real life testing not benches imo. Doing benches like this show maximum throughput which leans on those little cores from Intel a lot. We shall have that soon enough.

#6047072

winning.exe

2022-08-31 20:55

JamesSneed:

What matters is real life testing not benches imo. Doing benches like this show maximum throughput which leans on those little cores from Intel a lot. We shall have that soon enough.

Comments like this make me wonder 😱. My PC is used mainly for highly threaded code compilation, CAD, and rendering, with only occasional gaming. This is entirely throughput based and scales nearly linearly up to 128 threads (in my testing). If I had to buy a consumer processor, I would benefit greatly from a huge number of E-cores. Is this not a “real life” scenario 😀?

#6047083

kanenas

2022-08-31 21:13

this score is manual overclocking (INTEL) vs stock (AMD)

#6047085

nizzen

2022-08-31 21:25

kanenas:

this score is manual overclocking (INTEL) vs stock (AMD)

Score has no useful info, so it's useless.... Garbage at best 🙂

#6047090

TLD LARS

2022-08-31 21:43

winning.exe:

Comments like this make me wonder 😱. My PC is used mainly for highly threaded code compilation, CAD, and rendering, with only occasional gaming. This is entirely throughput based and scales nearly linearly up to 128 threads (in my testing). If I had to buy a consumer processor, I would benefit greatly from a huge number of E-cores. Is this not a “real life” scenario 😀?

Yes and no. The 5950x was mostly a better rendering CPU then 12900k if power usage was not ignored. The doubling of e-cores is going to give a good boost on the 13900k, but it still looks like the super fast p-cores are hungry at full load. I am still missing a lot of info on the 13900k I am afraid that everyone has the super turbo 350W setting enabled to inflate scores.

#6047092

nizzen

2022-08-31 22:02

TLD LARS:

Yes and no. The 5950x was mostly a better rendering CPU then 12900k if power usage was not ignored. The doubling of e-cores is going to give a good boost on the 13900k, but it still looks like the super fast p-cores are hungry at full load. I am still missing a lot of info on the 13900k I am afraid that everyone has the super turbo 350W setting enabled to inflate scores.

And in gaming, 12900k has 40-50% more fps per watt than 5950x (with fast memory) . Pretty funny it's almost the other way around in rendering. 🙂

#6047142

thestryker

2022-09-01 05:57

Tom's Hardware had an article about this and pointed out that the crypto test in Geekbench uses AVX512 which is why the scores here are so inflated. No test is perfect, but this one is particularly flawed.

#6047143

Kaleid

2022-09-01 06:41

That's a hefty single core performance raise. Bravo.