• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

*** AMD "Zen 4" thread (inc AM5/APU discussion) ***

g67575 said:
Is it safe to assume that Zen 4 will have more L3 cache perk core than Intel's 13th gen? Alder Lake was 2.5MB per core (for the P-Cores).
Click to expand...
intel-raptor-lake-s-design-cache-1024x600.png
 
Cheers. So, it looks like Intel has been lagging behind in L3 cache per core, for two Zen generations (including Zen 4).

Thing is, I might actually be tempted by Intel CPUs, if they had a CPU with a lot of L3 cache, like the 5800X3D. This CPU will probably be good for 3+ years for gaming, due to the large increase in minimum framerate. Shame it's expensive (for now), but not surprising for a new packaging technology.

I suppose that AMD has found a niche in the market, that is worth exploiting.

Technically, it looks like the 13th gen cores are still using the Goldencove spec, which supports upto 3MB of cache per core (for servers, less for 12th gen desktop). The same is true for the L2 cache, with Goldencove server chips supporting upto 2MB per core...
Link here:
en.wikipedia.org

Golden Cove - Wikipedia

en.wikipedia.org en.wikipedia.org

There doesn't seem to be an inherent limit on the number of E-Cores, within the Golden Cove spec.
 
videocardz.com

AMD EPYC "Zen4" Genoa CPU is 17% faster than "Zen3" Milan in single-core test - VideoCardz.com

AMD EPYC Genoa CPUs have been tested in Geekbench A dual-socket system featuring AMD upcoming Zen4 data-center CPU debuts on Geekbench benchmark charts. Two EPYC processors featuring a total of 192 cores and 384 threads have been spotted alongside Suma 65GA24 motherboard. This leak does not...
videocardz.com videocardz.com

Not a bad single-core geekbench result - 1460

Zen 4 could end up being ahead of the 12900KS in single core performance, if clocked >5ghz.
 
g67575 said:
videocardz.com

AMD EPYC "Zen4" Genoa CPU is 17% faster than "Zen3" Milan in single-core test - VideoCardz.com

AMD EPYC Genoa CPUs have been tested in Geekbench A dual-socket system featuring AMD upcoming Zen4 data-center CPU debuts on Geekbench benchmark charts. Two EPYC processors featuring a total of 192 cores and 384 threads have been spotted alongside Suma 65GA24 motherboard. This leak does not...
videocardz.com videocardz.com

Not a bad single-core geekbench result - 1460

Zen 4 could end up being ahead of the 12900KS in single core performance, if clocked >5ghz.
Click to expand...

If those numbers are right, 3.51Ghz Zen 4 vs 3.53Ghz Zen 3 then yes, that's actually about an 18% IPC gain. Much higher than expected.

If that translates to R23, and i'm not suggesting it will, i don't know, a 5.0Ghz Zen 4 should score around 1,970, the same as a 12900K.
5.7Ghz is 14% higher, so 2,240.

I think that would be pretty good, i was, and if i'm honest still am only expecting the 7950X to score around <2,100, i was expecting to see the 13900K to score about 10% higher ST, and i don't think that matters, at all

CPU Monkey has an unverified 13900K scoring 2,267.

On top of that if AMD could get them to boost to 6.0Ghz ST, it would score 2.360, that would frustrate Intel, but i am just being silly here...

https://www.cpu-monkey.com/en/cpu-intel_core_i9_12900ks
 
Essentially, I think the single core scores of Zen 4 and Goldencove will be almost the same, at the same clockrate. So, in general, I think the fastest CPUs (in most tasks) will be the ones with the highest all core frequencies* (for example in games).

In general, I think Zen 4 will achieve higher clockrates at lower cost in power, you won't need expensive CPU cooling to reach speeds >5ghz.

I don't think there will be a meaningful performance difference between the 12900KS and (most of the) 13th generation. The 13th gen i9s will probably have a small advantage in cache heavy tasks though.

Intel will almost undoubtedly win in multithreaded performance, due to the number of E-cores on the i9 processors.

*Actually, someone could test this theory, by running a 12900K or KS CPU at 3.5ghz in Geekbench, then checking the single core score.
 
g67575 said:
Essentially, I think the single core scores of Zen 4 and Goldencove will be almost the same, at the same clockrate. So, in general, I think the fastest CPUs (in most tasks) will be the ones with the highest all core frequencies* (for example in games).

In general, I think Zen 4 will achieve higher clockrates at lower cost in power, you won't need expensive CPU cooling to reach speeds >5ghz.

I don't think there will be a meaningful performance difference between the 12900KS and (most of the) 13th generation. The 13th gen i9s will probably have a small advantage in cache heavy tasks though.

Intel will almost undoubtedly win in multithreaded performance, due to the number of E-cores on the i9 processors.

*Actually, someone could test this theory, by running a 12900K or KS CPU at 3.5ghz in Geekbench, then checking the single core score.
Click to expand...

According to this the 12900K scores 1,893 ST, with a boost clock of 5.3Ghz.

Genoa (Zen 4) scores 1,460 at 3.5Ghz.

5.3 / 3.5 = 1.514 (51%)

1.460 + 51% = 2,204.

2.204 / 1,853 = 1.189 (+19%)

Make of that what you will, i'm inclined to think Geekbench doesn't score linier with clock speeds like that, or there is some other reason that's not right, because if it is its not just sandbagging, its taking the pee.
 
humbug said:
According to this the 12900K scores 1,893 ST, with a boost clock of 5.3Ghz.

Genoa (Zen 4) scores 1,460 at 3.5Ghz.

5.3 / 3.5 = 1.514 (51%)

1.460 + 51% = 2,204.

2.204 / 1,853 = 1.189 (+19%)

Make of that what you will, i'm inclined to think Geekbench doesn't score linier with clock speeds like that, or there is some other reason that's not right, because if it is its not just sandbagging, its taking the pee.
Click to expand...
I have to call BS on that link. Just tested it at 5.1 ST speed.

Micro-Star International Co., Ltd. MS-7D28 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7D28 with an Intel Core i9-12900K processor.
browser.geekbench.com
 
humbug said:
You're using faster memory than both, is what i take from this, again, sure.

Its a ll pretty academic, IMO there is something missing here, i'm about 80% sure we are not going to see Zen 4 cores scoring 2,200 at 5.3Ghz.
Click to expand...
Not sure what memeory they are using on your links but yeah, most likely im using faster. I stopped looking at gb numbers back when I boosted my 10900k's score by 38% just by going from 3200c16 to 4400c16 memory. Granted, that was GB4 which is even more memory reliant than GB5, but they are both pretty bad representatives of real world applications. Check the m1 numbers for example, they are scoring pretty high cause of fast memory access.
 
It sounds like we'd need someone to clock their 12900K /KS @ 3.5ghz, then clock the RAM @4800 MT/s. Then the Geekbench score would be a relatively fair comparison.

Edit - Maybe not actually, because I don't think we know what memory speed was being used with the Genoa CPU.
 
Clock speed is like rpm on a motor. You don’t lower the rpm of a flat plane crank V8 to match the lower rpm ceiling of a push rod for some silly “parity” comparison.

That’s just not how any of this works. Different architectures scale and peak differently.
 
You must log in or register to reply here.
Back
Top Bottom