New Intel Core i9-12900K sample spotted with DDR5-8000 RAM

nanoguy

Posts: 1,355   +27
Staff member
In brief: We're only a few weeks away from Alder Lake's official reveal, so it should be no surprise that more leaks are surfacing online. The latest one seems to suggest that Intel's 12th generation CPUs will have a robust memory controller that will allow enthusiasts to push DDR5 memory higher than DDR4 has ever gone.

We've seen a string of Alder Lake leaks as of late, but most were related to how well Intel's upcoming Core i9-12900K CPU performs against AMD's Ryzen 9 5950X, which is the current performance king. The Core i9-12900K, at least in its early incarnations, seems to be the best desktop part to come out of Team Blue's lab in a while, but there are still too many unknowns left to make a clear judgment about its potential.

There's been a lot of speculation around DDR5 and how it will influence Alder Lake's performance in specific scenarios such as gaming. The new CPUs will support both DDR4 and DDR5, and most of the leaks we've seen so far have been from a Core i9-12900K paired with DDR5 memory.

If a new leak from Twitter user REHWK is valid, Alder Lake CPUs will come with a robust memory controller that will be a gift to enthusiasts who love to take RAM to the extreme. In the CPU-Z screenshot shared by the leaker, we also get a further confirmation for the general specifications of the Core i9-12900K, a 16-core, 24-thread part with a TDP rating of 125 watts and no AVX-512 support.

In this new leak, the Alder Lake part appears to be paired with Gigabyte's DDR5-6200 memory kit on a Z690 Aorus Tachyon motherboard. By default, the DDR5-6200 memory will operate at DDR5-4800 speeds as per the JEDEC spec, with 42-39-39-77-116 timings at 1.1V. Under the XMP-6200 profile, timings are set a bit tighter to 38-38-38-76-125, but the operating voltage is bumped to 1.5V for stable operation. Otherwise, there's another XMP-6400 profile with more relaxed timings and an operating voltage of 1.45V.

To put things in context, an interesting feature that debuted with Rocket Lake is Gear modes, which are Intel's response to AMD's Infinity Fabric modes for memory overclocking. Gear 1 allows you to run the memory controller and the memory itself at the same frequency, while Gear 2 and Gear 4 allows you to run them at different frequencies. In the case of Gear 2, the memory controller runs at half the frequency of the memory, and the latter mode allows the memory controller to run at one-quarter of the memory frequency.

Alder Lake is also expected to have this feature, and this new leak suggests the memory controller on the new CPU lineup allows for ups to a 67 percent memory overclock to DDR5-8000 from the DDR5-4800 spec, with timings set to 50-50-50-100-150 for CL, tRDC, tRP, tRAS, and tRC, respectively. Notably, the overclocker achieved this using Gear 2 mode, meaning the memory controller is running at around 2,000 MHz, while the memory runs at 4,000 MHz (8,000 MHz effective speed). As of writing, the DDR4 overclocking record sits at 7,156 MHz, so this leak suggests that DDR5 should be capable of reaching higher than that right from the get-go.

Overall, it's an impressive feat that likely required some exotic cooling to pull off. One shouldn't read too much into it, as we have yet to see exactly how much of a speed benefit DDR5 brings over DDR4 and whether or not it'll be worth the price premium you'll have to pay for the privilege. At least one manufacturer plans to go beyond the 10,000 MHz barrier with overclocked modules, so it'll be interesting to see what DDR5 has in store.

Permalink to story.

 
Based on personal experience and Silicon Lottery closing up shop and, I've convinced myself to not bother with K and X SKU's sold for overclocking going forward until 500MHz+ overclocks are possible again on air or AIO.

As for ADL, bring on the reviews!
 
I saw the title to the article out of the corner of my eye and read it as "ddr2-800" which would have been WAY more interesting for, uhh, reasons....
 
Last edited:
That's 128Gb/s on a CPU with dual channel. Very impressive.
But those latency numbers are a bit high. That's a 12.5 ns real latency.
 
I saw the title to the article put of the corner of my eye and read it as "ddr2-800" which would have been WAY more interesting for, uhh, reasons....
My Prince, I have DDR-2 800 Mhz in this machine, which I use to fervently sing your praises..

I feel, of course, that you are deserved, at minimum, of DDR-4. But alas, my typing skills are not up to the task of blowing sunshine at such breakneck speeds..
 
Next year I will move to the new Zen AMD will release. Never interested in Intel really. And DDR5... I don't care.
 
My Prince, I have DDR-2 800 Mhz in this machine, which I use to fervently sing your praises..

I feel, of course, that you are deserved, at minimum, of DDR-4. But alas, my typing skills are not up to the task of blowing sunshine at such breakneck speeds..
I'm so lucky to have such a loyal servant, I think a night with the royal goat is in your future
 
Last edited:
1.5v on DDR5 no mention of the temps and active cooling to keep it down. DDR4 at 1.5 and above needs active cooling.
 
I'll be ready to build a new system in 2-3 years once DDR5 is widespread and the 13900K is available.
Yeah alot of people are saying in the next couple years it will be time, and by then Intel and AMD should have some really nice hardware out and maybe be available to buy for MSRP.
 
You don't need any of that. Heatsinks on RAM modules are for show, mostly.
Nope they talked about it already, ddr5 will be moving a micro controller on each stick so inevitably now you will need the heatsinks as they will generate alot of heat normally ram chips don't need cooling.
 
I'll be ready to build a new system in 2-3 years once DDR5 is widespread and the 13900K is available.

Well in 3 years you'll be buying Luna Lake and that is going to have about 60% IPC uplift over Alder Lake and revolutionary new cores, Intel;'s biggest architectural change according to insiders. Even Meteor Lake will be a 25% IPC uplift. Why anyone would rush out for Alder Lake and overpriced glacially slow DDR5 is beyond belief.

I think Meteor Lake vs Zen4 will be very interesting and both releasing in about 12 months.
 
I think Meteor Lake vs Zen4 will be very interesting and both releasing in about 12 months.
What I said earlier abour AMD fans should, "fear that they have awakened a sleeping giant", still stands.

For all the snot being blown up about Intel's profit taking instead of innovation, bear in mind, Intel has the capital to finance several of their own fabs, while AMD still has to grovel at TSMC's feet, to get a single CPU out the door.
 
Don't get too excited with DDR5, with how the GPU market is it will be the bottleneck for the DDR5.
 
Well in 3 years you'll be buying Luna Lake and that is going to have about 60% IPC uplift over Alder Lake and revolutionary new cores, Intel;'s biggest architectural change according to insiders. Even Meteor Lake will be a 25% IPC uplift. Why anyone would rush out for Alder Lake and overpriced glacially slow DDR5 is beyond belief.

I think Meteor Lake vs Zen4 will be very interesting and both releasing in about 12 months.
Even if those figures are right, what would be power consumption? 60% IPC improvement is very easy to accomplish. Keeping power consumption on reasonable levels same time is totally another question.

Only real reason for "hybrid" approach on Alder lake is power consumption.
What I said earlier abour AMD fans should, "fear that they have awakened a sleeping giant", still stands.

For all the snot being blown up about Intel's profit taking instead of innovation, bear in mind, Intel has the capital to finance several of their own fabs, while AMD still has to grovel at TSMC's feet, to get a single CPU out the door.
AMD's advantage is that they could use any manufacturer, except Intel of course. While they may partially use TSMC too, mostly Intel is stuck with their own fabs.

Apple have money to make everything themself. Still, they actually outsource almost everything. Why? Because Apple can buy from anyone that makes "best" stuff.

As Intel makes everything itself, Intel is also in trouble if something goes wrong. As we have seen.
 
Nope they talked about it already, ddr5 will be moving a micro controller on each stick so inevitably now you will need the heatsinks as they will generate alot of heat normally ram chips don't need cooling.

Still, the existing heatsinks they put on RAM sticks should suffice, most non-super-extra-expenssive motherboards have the whole RAM VRM section without any heatsinks over them and it's fine.
We shall see.
edit: you can look at the existing ddr5 on sale and there is no active cooling, just regular heatsinks.
 
Even if those figures are right, what would be power consumption? 60% IPC improvement is very easy to accomplish. Keeping power consumption on reasonable levels same time is totally another question.

Only real reason for "hybrid" approach on Alder lake is power consumption.

AMD's advantage is that they could use any manufacturer, except Intel of course. While they may partially use TSMC too, mostly Intel is stuck with their own fabs.

Apple have money to make everything themself. Still, they actually outsource almost everything. Why? Because Apple can buy from anyone that makes "best" stuff.

As Intel makes everything itself, Intel is also in trouble if something goes wrong. As we have seen.
Easy to say 60% IPC performance increase is very easy to accomplish but I call BS. Intel hhasn't cared about power for several years and could only deliver about 10% IPC uplift in 4 generations. Also AMD is saying the exact same thing as Intel. By 2025 both will be looking at 70% IPC uplift over current gen CPU's. Do you think they'll just clock the buggery out of these CPU's. Luna Lake is an entirely new architecture.

Intel has snapped up TSMC 3nm node, AMD will find it hard to get volumes they need for Zen 5 unless many new fabs are online or Samsung actually gets competitive. I'd like tyo know who all these other Fabs AMD can use. Intel muscling in on TSMC should be a worry for AMD.

And in case AMD fanboys get their nickers in a knot, I am wholly in on AMD and have said many times I won't be buying back into Intel for a long time. But I welcome strong competition and Alder Lake is just the start, it's be far the weakest CPU range over the next 3 years. Zen 4 and Zen 5 will need to step up a lot in performance IMO as they won't be smashing Intel any more going forward.
 
Easy to say 60% IPC performance increase is very easy to accomplish but I call BS. Intel hhasn't cared about power for several years and could only deliver about 10% IPC uplift in 4 generations. Also AMD is saying the exact same thing as Intel. By 2025 both will be looking at 70% IPC uplift over current gen CPU's. Do you think they'll just clock the buggery out of these CPU's. Luna Lake is an entirely new architecture.
At least current Intel CPU's work without water cooling being mandatory. +60% IPC is very easy if thermals and/or clock speeds could be ignored. Overall CPU would be much worse but I didn't say that...

Luna Lake being totally new architecture and still being Lake-series makes no sense.
Intel has snapped up TSMC 3nm node, AMD will find it hard to get volumes they need for Zen 5 unless many new fabs are online or Samsung actually gets competitive. I'd like tyo know who all these other Fabs AMD can use. Intel muscling in on TSMC should be a worry for AMD.
Intel gets small percentage of TSMC 3nm node. Like I commented, that majority was fake news. TSMC total 3nm capacity 10K wafers :joy:
And in case AMD fanboys get their nickers in a knot, I am wholly in on AMD and have said many times I won't be buying back into Intel for a long time. But I welcome strong competition and Alder Lake is just the start, it's be far the weakest CPU range over the next 3 years. Zen 4 and Zen 5 will need to step up a lot in performance IMO as they won't be smashing Intel any more going forward.
Zen4 details are unknown as usual. AMD no longer gives info like they did with Zen. But Intel has right now much more to catch up.
 
AMD's advantage is that they could use any manufacturer, except Intel of course. While they may partially use TSMC too, mostly Intel is stuck with their own fabs.
So why then, if Intel is "stuck with their owns fabs Al least according to you, are all the AMD fanboys skirting the issue of Intel breaking ground on 3 or 4 new fabs.

I suppose they'll ostensibly be, "stuck" with those as well./

BTW, didn't Intel bid successfully on TSMC's future 5 nm production? Or did we conveniently miss that blurb?"
 
Back