It's been a long wait, but "maybe" this Winter Intel will have a winner.

[RCHeliguy]

"IF" this leak is real, than the new 10nm i9-12900K which should be released at the end of the year would have a single core performance that is 62% better than my i9-9900K. My i9-9900K has a 500 vs 810 for single core in the R20 3D rendering benchmark. With PCI Express 5.0 and DDR5 RAM, it looks like enough of an upgrade to be worth it.

Leaked Intel Alder Lake numbers place it up to 22% quicker than AMD's finest Ryzen

Intel's gamble on big.LITTLE architecture could be about to pay off big time.

www.pcgamer.com

Fortunately it looks like DDR5 memory is in production despite the lack of motherboards that can make use of them, but as this article says, people appear to be buying them up in anticipation. They appear to initially be DDR5 32Gb x 4800 MHz.

The first 32GB DDR5 RAM kit lands at retail for $311 and it’s already sold out

TeamGroup's first consumer DDR5 memory has arrived, there is just nowhere to install it (yet).

www.pcgamer.com

These rumors say we could start seeing these chips in October and Intel will focus on the higher end K and KF CPU's until 2022. So we should have CPU's and DDR5 memory, and then we have to wait for motherboard availability from China.


It looks like there will be an interim release in 2023 and it could be another 15-20% performance increase, but the next BIG upgrade looks to be 2024 once again "IF" Intel gets their ribbonFET's into production by then. If they do then even with 7nm lithography they would get closer to a 2nm transistor density. That should be a pretty big jump.

 

[Neilski]

"IF" this leak is real, than the new 10nm i9-12900K which should be released at the end of the year would have a single core performance that is 62% better than my i9-9900K. My i9-9900K has a 500 vs 810 for single core in the R20 3D rendering benchmark.

That's an insanely large increase and I was initially very sceptical.
However, googling around a bit, I found data which suggests that Cinebench (or at least R20 single core, haven't looked much further yet) speeds do vary hugely even between say the 10900K and the 10700K, by massively more than the boost clock variance. Maybe it's very sensitive to cache size. Even the 11900K is already more than 40% quicker than your 9900K, for that particular benchmark. IMHO this is highly unlikely to be matched by the performance in a sim, but only testing will answer that.

Conclusion might be that maybe you really should expect a 62% leap in single-core performance over your "old" CPU IFF you spend a lot of time running Cinebench R20 in single core mode
For more normal workloads I'd be expecting maybe a ~25% improvement, which is certainly not to be sniffed at of course.

Spoiler: Some example data to back it up

Cinebench R20 single core data from www.cpu-monkey.com:
11900K = 732
11700K = 612
=> the i9 is 20% faster
10900K = 645
10700K = 524
=> the i9 is 23% faster
and 11th gen i9 is faster than 10th by 13%

Passmark single thread ratings from www.cpubenchmark.net:
11900K = 3493
11700K = 3473
=> the i9 is 0.6% faster
10900K = 3164
10700K = 3082
=> the i9 is 2.7% faster
and 11th gen i9 is faster than 10th by 10%

Hmm.

 

[RCHeliguy]

I'll be waiting for the actual gaming benchmarks.

I'm currently running 3200MHz DDR4 memory and I'm hoping 4800MHz DDR5 memory will help too, but I understand that the memory timing boost will be much smaller than the CPU itself.

 

[Deleted member 197115]

You can still get 4400Mhz DDR4 that works on Z390
E.g.

F4-4400C18D-16GTZR-G.SKILL International Enterprise Co., Ltd.

G.SKILL

www.gskill.com

I am running 4133MHz variant of the same DIMM with i9-9900K.

 

[RCHeliguy]

I'm not going to bother upgrading my existing MB at this point. I'll just move it to my development box whenever the next big enough difference arrives.

 

[Deleted member 197115]

I'm not going to bother upgrading my existing MB at this point. I'll just move it to my development box whenever the next big enough difference arrives.

Why would you need to upgrade it, you are on Z390 chipset?
What MB do you use currently, try G-Skill configurator to see which stick works for you specific MB.

 

[RasmusP]

That's an insanely large increase and I was initially very sceptical.
However, googling around a bit, I found data which suggests that Cinebench (or at least R20 single core, haven't looked much further yet) speeds do vary hugely even between say the 10900K and the 10700K, by massively more than the boost clock variance. Maybe it's very sensitive to cache size. Even the 11900K is already more than 40% quicker than your 9900K, for that particular benchmark. IMHO this is highly unlikely to be matched by the performance in a sim, but only testing will answer that.

Conclusion might be that maybe you really should expect a 62% leap in single-core performance over your "old" CPU IFF you spend a lot of time running Cinebench R20 in single core mode
For more normal workloads I'd be expecting maybe a ~25% improvement, which is certainly not to be sniffed at of course.

Spoiler: Some example data to back it up

Cinebench R20 single core data from www.cpu-monkey.com:
11900K = 732
11700K = 612
=> the i9 is 20% faster
10900K = 645
10700K = 524
=> the i9 is 23% faster
and 11th gen i9 is faster than 10th by 13%

Passmark single thread ratings from www.cpubenchmark.net:
11900K = 3493
11700K = 3473
=> the i9 is 0.6% faster
10900K = 3164
10700K = 3082
=> the i9 is 2.7% faster
and 11th gen i9 is faster than 10th by 10%

Hmm.

Are you sure that cpu monkey is a reliable source for this?
From the cinebench results I'm remembering, the i9s were only faster due to higher boost clocks but not much apart from what made sense.

Intel Core i9-11900K processor review

We review Intel's new Desktop flagship processor, the 8-core Core i9-11900K. The 11th generation Intel Core desktop processors previously known as RKL aka Rocket Lake-S has been released in an effort to take on the battle with AMD. Performance - CineBench 15 (and IPC)

www.guru3d.com

I can recommend guru3d for a good overview and mostly stock clocks. They also do some oc runs but don't put them into the overview charts (sadly).

All CPUs locked to 3.5 GHz barely show any difference going from i9 to I7 to i5 within one generation!

R20 results:

Intel Core i9-11900K processor review

We review Intel's new Desktop flagship processor, the 8-core Core i9-11900K. The 11th generation Intel Core desktop processors previously known as RKL aka Rocket Lake-S has been released in an effort to take on the battle with AMD. Performance - CineBench 20

www.guru3d.com

Sadly no IPC chart but you can see that it's mostly clock speed and generation related. The 10th and 9th Gen are very very close in single thread performance as expected.

However the internal latencies of the 11th Gen really hits at high fps in games...
It seems to be better at lower fps like this one polish acc benchmark shows:

Test procesorów Intel Core i5-11600K vs AMD Ryzen 5 5600X vs Intel Core i5-10600K. Które sześć rdzeni jest najlepsze? | PurePC.pl

Test procesorów Intel Core i5-11600K vs AMD Ryzen 5 5600X vs Intel Core i5-10600K. Które sześć rdzeni jest najlepsze? (strona 60) Test procesorów Intel Core i5-11600K vs AMD Ryzen 5 5600X vs Intel Core i5-10600K. Które sześć rdzeni jest najszybsze w grach i programach?

www.purepc.pl

 

[Neilski]

Are you sure that cpu monkey is a reliable source for this?

Nope! I'm very much a novice at finding Cinebench data

All CPUs locked to 3.5 GHz barely show any difference going from i9 to I7 to i5 within one generation!

Ahh, interesting. Seems like the other data is... questionable then.

 

[RasmusP]

Nope! I'm very much a novice at finding Cinebench data

Ahh, interesting. Seems like the other data is... questionable then.

Rule Nr 1: all pages that don't to the actual reviews can't be trusted

CPU mark is a great way to do quick comparisons but sometimes questionable too.
It's my go to page for stuff like comparing a laptop cpu from 2013 with a modern Ryzen 3300€ whole throwing in my old 2600k in the mix.

But for modern only charts, guru3d is a great page!
The other go to site is gamersnexus. But mostly you have to skip through the videos (which are nicely structured!) to get the info.

 

[RCHeliguy]

Why would you need to upgrade it, you are on Z390 chipset?
What MB do you use currently, try G-Skill configurator to see which stick works for you specific MB.

You misunderstand me. I don't want to upgrade anything on my MB, because it seems like a waste.

I'm just going to pass the MB with CPU, and memory to my development machine whenever I upgrade my gaming computer. My dev boxes i7-2600K is still plenty fast enough and the i9 will be overkill, but will allow a Win 11 update that I'll need for development purposes and the i9 will be nice when rendering video once in while.

I may even wait to do any upgrades until I'm planning a new VR headset, MB, CPU, GPU, and memory.

 

[HoiHman]

That's an insanely large increase and I was initially very sceptical.
However, googling around a bit, I found data which suggests that Cinebench (or at least R20 single core, haven't looked much further yet) speeds do vary hugely even between say the 10900K and the 10700K, by massively more than the boost clock variance. Maybe it's very sensitive to cache size. Even the 11900K is already more than 40% quicker than your 9900K, for that particular benchmark. IMHO this is highly unlikely to be matched by the performance in a sim, but only testing will answer that.

Conclusion might be that maybe you really should expect a 62% leap in single-core performance over your "old" CPU IFF you spend a lot of time running Cinebench R20 in single core mode
For more normal workloads I'd be expecting maybe a ~25% improvement, which is certainly not to be sniffed at of course.

Spoiler: Some example data to back it up

Cinebench R20 single core data from www.cpu-monkey.com:
11900K = 732
11700K = 612
=> the i9 is 20% faster
10900K = 645
10700K = 524
=> the i9 is 23% faster
and 11th gen i9 is faster than 10th by 13%

Passmark single thread ratings from www.cpubenchmark.net:
11900K = 3493
11700K = 3473
=> the i9 is 0.6% faster
10900K = 3164
10700K = 3082
=> the i9 is 2.7% faster
and 11th gen i9 is faster than 10th by 10%

Hmm.

That does indeed look very good and might the right upgrade for my 9700k.

The issue with all these comparisons is that they usually compare stock to stock performance. But who has his 9700k/9900k running on stock speeds nowadays?

Mine runs at 5Ghz, but still the improvement should be significant.

 

[RasmusP]

The issue with all these comparisons is that they usually compare stock to stock performance. But who has his 9700k/9900k running on stock speeds nowadays?

More people than you'd think...
My favorite though are the comparisons with the i5 k's, which clock quite a lot lower than the i7 and i9...
They look quite bad in all benchmarks, until they looks very good when all cpu's are overclocked.
Sure, the i9 10900k boosts to 5.3 on single cores and my i5 10600k can only run stable at 4.9 GHz all core.
But the difference when you put both on the highest stable oc is very close, while the price difference is insane for "just gaming".

For other tasks, you need the cores though. My work applications are single thread only or run on the graphics card...

 

[RCHeliguy]

Hmmm....

 

[Deleted member 197115]

Hmmm....

Nice

 

[RCHeliguy]

Of course these are the same numbers I mentioned initially just in video form.

But I love his enthusiasm

 

[RCHeliguy]

Samsung is releasing 7200MHz DDR5 memory with up to 512Gb per board for servers and is probably very expensive.

Pretty impressive stuff! But there is no word if Samsung will release consumer memory in smaller sizes for gamers. It would be great if they released 32 and 64Gb sizes. Whether gaming MB's will support this speed is the next question.

 

[RCHeliguy]

The next Intel CPU's and motherboards should launch November 19th.

I imagine the best well optimized motherboards will come a bit later.