Intel Core i9-12900K Single-Thread Performance Allegedly 27% Faster Than Ryzen 9 5950X

[R0H1T]

What if ADL forces AMD to push forward the zen4 launch ~ oh wait you think that's impossible? Zen w/3D V-Cache could well just be a filler with not an entire stack full of that absurd cache amount! You're just clutching at straws with one speculation after another

 

[Crackong]

8 cores + 8 little/fake cores. What a joke Intel.


Intel can't get a true 16 core working so they create 8 big core and 8 little cores. What a joke.

Getting true 16 core is way too expensive (and also HOT, very HOT) for Intel MSDT architecture.

If Ponte vecchio actually works out, we might see Intel MSDT CPUs with MCM.

But before that? very unlikely

 

[Hossein Almet]

My 5950X single score is 688 points.

 

[Fouquin]

I'm reminded of when Zen 1 "beat" Kaby Lake in the CPU-Z single core test, and how that didn't play out in any other scenario.

 

[Vya Domus]

Since it's still not clear whether this is on DDR5 or DDR4 all of these benchmarks are worthless, because just faster memory alone can boost performance by a lot in a CPU.

Getting true 16 core is way too expensive

Where did you get that from ? It's too expensive if you want colossal margins maybe, like Intel usually wants. But otherwise there is nothing "too expensive" about 16 cores, even on a monolithic die. After all, we've been told their 7nm node is amazing.

 

[AnarchoPrimitiv]

That is impressive. Bring on the next gen HEDT with these designs already!

This is impressive compared to Intel's precious blunders... Given that they have an R&D budget literally more than 6x greater than AMD's, this is simply what they should be doing and should HAVE been doing. That said, I have been claiming that as consumers, with respect for the long view of things, it would have be more beneficial for us as consumers if AMD had a few more years (at least 3 to 5 additional years) of Dominance in order to build up their financial resources and better penetrate the segments with the largest T.A.M.'s within the x86 market as a whole: enterprise and mobile. Name another industry in which a company like AMD (2020 R&D budget of $1.98 Billion) has been able to truly compete (as in NOT being relegated to only playing the role of the perennial budget offering for which consumers do not willingly choose, but merely settle) against a monopolist like Intel (2020 R&D budget $13.56 Billion) for a decade or longer and for the same markets and market segments.... I'm having trouble thinking of any.

Preliminary reports, including ones printed on this very website have claimed Zen4 to have an IPC increase of 29%, funny. Perhaps with V-cache, frequency bumps and some other architectural tricks, core for core performance uplift could be 35%+, but it seems as though AMD shouldn't be waiting around much longer. The past has proven that Lisa Su will do the absolute best with what she is given, her compitence isn't in question, but what is is whether she has access to enough resources to compete with a company whose financial power is a magnitude greater and has established a mindshare so pervasive that the vast majority of consumers (who are NOT well educated with respect to hardware specifications and empirical data from the latest reviews) have the terms "Intel" and "laptop" synonymous in their minds to such a degree that the thought of even inquiring whether there is a competing option never even occurs to them.

 

[robb]

What if ADL forces AMD to push forward the zen4 launch ~ oh wait you think that's impossible? Zen w/3D V-Cache could well just be a filler with not an entire stack full of that absurd cache amount! You're just clutching at straws with one speculation after another

the fact that you think amd could magically bring forward the launch of zen4 by nearly a year shows you have no business even discussing anything tech related.

 

[R0H1T]

Who said anything about one year? Do you have that number hardwired in the brain currently?

 

[Melvis]

erm when does the 11900K beat the 5950X? or am I missing something here?

 

[olymind1]

I'm more curious about i5-12400 cpus, with cheaper chipset/motherboard and with more affordable DDR4 memory slots. How will those compare to Zen2-3s and to 10-11th Gen i5s..

 

[R0H1T]

Better with DDR5 than DDR4, the top mem speeds supported by locked (Intel) CPU is rather anemic so don't expect too much. I'd hope someone finally tests these with top end DRR4 & DDR5 memory, even if on different boards!

 

[Crackong]

Where did you get that from ? It's too expensive if you want colossal margins maybe, like Intel usually wants. But otherwise there is nothing "too expensive" about 16 cores, even on a monolithic die. After all, we've been told their 7nm node is amazing.

You've answered yourself.
Intel wants margin

AMD is selling 5950x at $749 right now.

If Intel really come up with a 16 cores monolithic ringbus
How much will it cost? >$1k ?

Will it be 33% faster than 5950x ? Probably not.

 

[R0H1T]

It won't cost a grand unless the yields are horrific, they're already selling the massive 117xx/9xx chips at a reasonably competitive price ~ wrt what Intel usually sells them at. I'd argue a 16 (large) core design would probably be at most 10-20% bigger than the 12900k.

Will it be 33% faster than 5950x ? Probably not.

Intel's still selling their ST performance & clock speeds, with 16 large cores they'll usually go down or power consumption shoots way way up! I doubt they wanted to compromise on the single core scores plus yields, otherwise 16c monolithic designs may have come out. I mean Sapphire Rapids is on 10nm, so this should not be a big deal if Intel wills it.

 

[Vya Domus]

Intel wants margin

That's their problem but to say that's it's too expensive doesn't mean anything. They can if they want to.

A 16 core die shouldn't be too costly on 7nm.

 

[stimpy88]

Yeah right, we will see in a short time. Until then, I call BS on these "leaks".

 

[docnorth]

Intel PR1 : Hey we must do something about that SiSoftware leak.
Intel PR2 : Here is a screen shot with no model numbers just the score, let's use it.

Big companies might indeed promote leaks that favor their products, but usually only if the results can be confirmed. If Intel provided some "motivation", then this leak is probably real.

 

[BorisDG]

While improvement is great and welcome, who the hell will need more performance? You can still play all new titles even with 6700K.

 

[Makaveli]

My 5950X single score is 688 points.

The reference systems they use for this are not optimized.

So the scores are always lower than on you will see from our machines.

I score 672 single core with my 5800X but i'm on Cas 14 and tuned memory timings.

 

[phanbuey]

Afaik -- the cost of the ring is not the main issue, it's that a ring with 16 cores + imc + secret sauce chip + io has way too much latency between the cores/ cache with that many stops... even a ring of rings at that size isn't the greatest when data needs to flow between the furthest cores. Mesh is too slow and doesn't make sense for consumer cpu, and 14nm tech makes it impossible to compete on the HEDT side in raw core counts vs AMD.

Given the hand they're playing with, seems like this is a great move.

This chip design makes a quite a bit more sense for the masses if they can get the scheduler and software to work properly on it -- a 5950x like the 3950x and the 1950x will get smoked by the next gen mainstream AMD x600x/x800x variants in games and the majority of consumer apps, so it won't really be surprising that Intel Alder Lake will be faster than a 5950x with exception of some heavily threaded tasks.

 

[zlobby]

Uhm, at what power, temps and clock?

 

[yotano211]

I get 637 on single thread and 5975 on a i7 11800h laptop processor running at 4.3ghz.

 

[Selaya]

[ ... ]
Nevertheless, I hope these results translate to gaming performance because I've been waiting a long time for this kind of performance jump from Intel. [ ... ]

Unfortunately 1T perf will do exactly jack for gaming performance, the games that don't scale with cores beyond 6 already cap out w/ the 5600X (ie, benches the same whether you use 5600X or 5950X) as all Vermeer parts have the same 32MiB of L3.
Alder Lake will cap out at 32MiB of L3 too, so it'll do about jack-all about improving gaming performance. (Obviously, a better IMC than the anemic GloFo one may help a bit, but.)
However. AMD will have the Vermeer refresh ready, featuring a massive 96MiB of L3. That'll totally destroy Alder Lake w/ regards to gaming, to the extent that 1T may actually matter again, but yeah.

 

[TheoneandonlyMrK]

I did say it's going to be close, but.

 

[Why_Me]

I'm more curious about i5-12400 cpus, with cheaper chipset/motherboard and with more affordable DDR4 memory slots. How will those compare to Zen2-3s and to 10-11th Gen i5s..

This ^^

 

[TheoneandonlyMrK]

This ^^

Not 29% at a guess, lower in fact.