Question Raptor Lake - Official Thread

[Hulk]

Since we already have the first Raptor Lake leak I'm thinking it should have it's own thread.
What do we know so far?
From Anandtech's Intel Process Roadmap articles from July:

Built on Intel 7 with upgraded FinFET
10-15% PPW (performance-per-watt)
Last non-tiled consumer CPU as Meteor Lake will be tiled

I'm guessing this will be a minor update to ADL with just a few microarchitecture changes to the cores. The larger change will be the new process refinement allowing 8+16 at the top of the stack.

Will it work with current z690 motherboards? If yes then that could be a major selling point for people to move to ADL rather than wait.

 

[Hulk]

Just to give you some first hand experience - I've got a 14900K and this ended up being my VID. I've got more money than sense and this is my hobby so I don't really care about the $600 for a minimal upgrade.

View attachment 87918

When comparing against my old 13900K (SP 96) this is what I've noticed.

Memory Stability at default IMC Voltage (64Gb Hynix A-Die dual rank):

13900K - Barely stable at DDR5-6600. The default IMC voltage was 1.389V. I mostly ran it at DDR5-6400 just for peace of mind and 100% stability.
14900K - Stable at DDR5-6800. The default IMC voltage is 1.368V. The same tests that would error within minutes at this speed now can run for hours at a time. It's been a week of multiple reboots / retrainings, stress tests and gaming with zero errors.

Clocks:

13900K - 1.449V VID for 5.8ghz (stable undervolt was -0.040/-0.020 for P & E cores respectively)
14900K - 1.443V VID for 6.0ghz (stable undervolt is -0.050/-0.0330 for P & E cores respectively... I could push it further but it's just not worth it since I've never seen a temperature exceed 83* C at 253W)

With both CPUs capped at 253W, I get ~140-180mhz more frequency at same power consumption and 10-12*C lower temps. I would advise most people with 14900K to just cap E-cores at 4.3ghz (it's what I did) since it lowers power consumption and enables higher boost on P-cores with the same PL2.

That's a nice CPU you have there. Under "Prediction P-Core V for 6000MHz" what do the 1.307 and 1.543 volts refer to. Line and droop or something like that? And thanks for posting your experience.

One more question. I have a 280 AIO that can comfortably handle about 225W. What P and E core frequency with CB R23 can you do at that power setting? Or what would you suggest?

 

[H433x0n]

That's a nice CPU you have there. Under "Prediction P-Core V for 6000MHz" what do the 1.307 and 1.543 volts refer to. Line and droop or something like that? And thanks for posting your experience.

As far as I know, I think it's related to Asus' AI overclocking feature (which I wouldn't recommend BTW). The first thing I do is disable Asus' MCE and select "Disabled - Enforce all limits". This will have your CPU running at Intel's default PL1/PL2 and target clock frequencies.

One more question. I have a 280 AIO that can comfortably handle about 225W. What P and E core frequency with CB R23 can you do at that power setting? Or what would you suggest?

I would say that if your AIO could handle 225W on an Alder Lake CPU, it could probably handle ~240-250W on RPL-R CPU without throttling. I can tell you on my system at 253W, I will average roughly ~5650mhz in the OCCT CPU stability test. Whereas this same test under the same conditions on my 13900K managed ~5450mhz at same power consumption. As for the ECore clocks, my 14900K will maintain a locked 4300mhz (I reduced multiplier in BIOS to 43x), my 13900K more or less could maintain 4300mhz as well.

 

[Hulk]

As far as I know, I think it's related to Asus' AI overclocking feature (which I wouldn't recommend BTW). The first thing I do is disable Asus' MCE and select "Disabled - Enforce all limits". This will have your CPU running at Intel's default PL1/PL2 and target clock frequencies.

I would say that if your AIO could handle 225W on an Alder Lake CPU, it could probably handle ~240-250W on RPL-R CPU without throttling. I can tell you on my system at 253W, I will average roughly ~5650mhz in the OCCT CPU stability test. Whereas this same test under the same conditions on my 13900K managed ~5450mhz at same power consumption. As for the ECore clocks, my 14900K will maintain a locked 4300mhz (I reduced multiplier in BIOS to 43x), my 13900K more or less could maintain 4300mhz as well.

Good info. Thanks.
I'd be happy with all-core P's at 5.4GHz and E's at 4.2GHz actually. I'll probably set a 30 second PL2 of 250W for bursty workloads.

 

[jpiniero]

Intel Core i9-14900K Raptor Lake Tested at Power Limits Down to 35 W

There's no denying that Intel's Core i9-14900K is a power hog. This article examines the power, performance and thermals effects of limiting its power consumption, with very interesting results. Additionally, we investigate undervolting to determine if efficiency can be optimized even further.

www.techpowerup.com

14900K power scaling test in games. At 125 W, the performance hit isn't much... but at 65 W it is, let alone 35 W.

 

[Hulk]

Out of boredom tonight I pulled the trigger on a 14900K. It'll be here Tuesday. I'm sure I'll get a crappy one as I always do.

 

[igor_kavinski]

I'm sure I'll get a crappy one as I always do.

Wishing you the rarest golden sample!

 

[H433x0n]

Out of boredom tonight I pulled the trigger on a 14900K. It'll be here Tuesday. I'm sure I'll get a crappy one as I always do.

Don’t worry about the binning too much. Even some of the worse bins are still better than the lower end 13900KS.

With proper bios configuration it’s a great processor - it’ll be fun to tinker with.

 

[Hulk]

Don’t worry about the binning too much. Even some of the worse bins are still better than the lower end 13900KS.

With proper bios configuration it’s a great processor - it’ll be fun to tinker with.

True. I've seen the best samples show 6GHz VID at 1.32 or so with the worst being at 1.5V. I'm not planning to run that high anyway so it not a huge deal. Would be fun to get a better one though.
I'm actually thinking of running with HT off, which I read distributes the heat more evenly on the die. Plus with 16 E's I'd rather have the threads not hitting the P go directly to the E's and avoid the Thread Director making errors in thread allocation.

 

[Hulk]

Ran a quick test. This is interesting.
13600K and Cinebench R23 MT.

6+6, no HT. P's at 5.1GHz, E's at 3.9GHz (stock)
Score is 17, 600, Core power 118W, max temp 64C

6+0 with HT, P's at 5.1GHz (stock)
Score is 15,000, Core power 116W, max temp 71C

So how do I interpret this? If you have an application that only uses 12 threads then the Thread Director would get it 100% correct if it did not send any threads to the logical cores, just 6 threads to the physical P cores, and 6 to the E's. Performance is better, efficiency is better, and temps are lower.

Of course as soon as you hit 13 or more threads then you'll want to use all of your available resources.

But with the 14900K there is some nuance to this. First of all you have 16 E cores available. That's a lot. There aren't a lot of apps as ridiculously multithreaded as CB. Second, since most Raptor 8+16 are thermally limited, by turning off HT you can achieve higher P core clocks and there are still many applications that need to feed on CPU cycles more than anything else.

Going to have to fool around with this when my 14900K comes in. I have noticed that when I'm mixing down an audio project in Presonus Studio One the render is faster with HT turned off. Since Studio One isn't well threaded the Thread Director is obviously making mistakes and sending threads to the logical cores rather physical ones. Just one example where running with HT off will be better for one of my use cases that often slows down my workflow.

 

[H433x0n]

Ran a quick test. This is interesting.
13600K and Cinebench R23 MT.

6+6, no HT. P's at 5.1GHz, E's at 3.9GHz (stock)
Score is 17, 600, Core power 118W, max temp 64C

6+0 with HT, P's at 5.1GHz (stock)
Score is 15,000, Core power 116W, max temp 71C

So how do I interpret this? If you have an application that only uses 12 threads then the Thread Director would get it 100% correct if it did not send any threads to the logical cores, just 6 threads to the physical P cores, and 6 to the E's. Performance is better, efficiency is better, and temps are lower.

Of course as soon as you hit 13 or more threads then you'll want to use all of your available resources.

But with the 14900K there is some nuance to this. First of all you have 16 E cores available. That's a lot. There aren't a lot of apps as ridiculously multithreaded as CB. Second, since most Raptor 8+16 are thermally limited, by turning off HT you can achieve higher P core clocks and there are still many applications that need to feed on CPU cycles more than anything else.

Going to have to fool around with this when my 14900K comes in. I have noticed that when I'm mixing down an audio project in Presonus Studio One the render is faster with HT turned off. Since Studio One isn't well threaded the Thread Director is obviously making mistakes and sending threads to the logical cores rather physical ones. Just one example where running with HT off will be better for one of my use cases that often slows down my workflow.

I’m pro removing hyperthreading for a desktop CPU. As soon as I’m done validating my mild undervolt - I’m turning it off and it’ll probably never be enabled again.

As far as thread director is concerned, it’s supposed to schedule PCore>ECore>PCore HT. To be honest, I haven’t noticed much of a benefit of it being off from a scheduling standpoint.

In many instances where hyperthreading is a performance benefit in a desktop environment you’re better off just offloading that workload to a GPU anyway since it’s a stupidly parallel task unfit for a CPU.

As far as clocks are concerned, absolutely you’re able to achieve higher clocks for longer with less heat and power. For any ST bound task, you’re almost always better off with HT disabled (this includes gaming).

 

[Hulk]

The last few posts kind of illustrate why Intel is rumored to be moving away from HT. The transistors used to implement HT in the P's can be more advantageously used to increase ST performance. Since the E's are so efficient from an area point of view 4 clusters of E's can handle the applications that do require more than 8 threads. Where Intel can run into problems (and does) with AMD is when an application utilizes between 9 and 16 threads. In this case Intel has to compete against 8 of the Zen cores with E cores.

 

[igor_kavinski]

Does turning HT off result in better IMC performance too, due to the lower heat output? Would the IMC be able to hit higher frequences or lower timings more easily with HT off?

 

[Hulk]

My 14900K came in today. What a piece of crap. 1.438VID for 5700MHz. Even set to 175 Watts P0 is hitting 100C with my 280 AIO with HT off! I'll check thermal contact but I'm sure this particular sample is a dog. It's going right back to Amazon without delay.

 

[igor_kavinski]

It's going right back to Amazon without delay.

I hope you try a KF next.

 

[H433x0n]

My 14900K came in today. What a piece of crap. 1.438VID for 5700MHz. Even set to 175 Watts P0 is hitting 100C with my 280 AIO with HT off! I'll check thermal contact but I'm sure this particular sample is a dog. It's going right back to Amazon without delay.

Can you post the VID? It doesn’t need to be a board that supports an SP score it should still display a VID. If you can’t find it in BIOS there is an Asus tool that will display it from within Windows (I could get the link for you).

 

[Hulk]

Can you post the VID? It doesn’t need to be a board that supports an SP score it should still display a VID. If you can’t find it in BIOS there is an Asus tool that will display it from within Windows (I could get the link for you).

Unfortunately I already pulled it. I don't want to waste time with it. That VID I posted was a BIOS readout. I am curious about the Asus VID tool if you can find a link?

Also while running CB R24 ST at 5400MHz the VID from HWinfo averaged 1.403, which is really high for that frequency.

 

[H433x0n]

Unfortunately I already pulled it. I don't want to waste time with it. That VID I posted was a BIOS readout. I am curious about the Asus VID tool if you can find a link?

Also while running CB R24 ST at 5400MHz the VID from HWinfo averaged 1.403, which is really high for that frequency.

Download the M15_OCTool from overclock.net in this thread. Inside that zip folder you'll find an app that's simply labeled "Tool.exe". Inside the "CPU" drop down select the option "VF points".

 

[Thunder 57]

My 14900K came in today. What a piece of crap. 1.438VID for 5700MHz. Even set to 175 Watts P0 is hitting 100C with my 280 AIO with HT off! I'll check thermal contact but I'm sure this particular sample is a dog. It's going right back to Amazon without delay.

Oh no, it runs within specs but not to my liking! Returning right away! I don't care about others who will have to absorb this cost!

 

[Hulk]

Oh no, it runs within specs but not to my liking! Returning right away! I don't care about others who will have to absorb this cost!

Can't really argue as I am being somewhat of a baby. But, even at stock it won't hold 6GHz ST during CB 23 run, and that's with my 280 AIO. So technically it is not running in spec. I want one that will run that bench in spec with a good AIO. Perhaps that is too much to ask for my $600.

 

[Hulk]

Download the M15_OCTool from overclock.net in this thread. Inside that zip folder you'll find an app that's simply labeled "Tool.exe". Inside the "CPU" drop down select the option "VF points".

Thanks.

 

[coercitiv]

Can't really argue as I am being somewhat of a baby. But, even at stock it won't hold 6GHz ST during CB 23 run, and that's with my 280 AIO.

Are you using a contact frame? Not saying you must be using one to get stock perf, but if you're going to try another 14900K then consider buying one if you haven't already.

 

[Hulk]

Are you using a contact frame? Not saying you must be using one to get stock perf, but if you're going to try another 14900K then consider buying one if you haven't already.

I am going to reinvestigate my 14900K today, give it a second chance.
Yes, I have a contact frame.

 

[igor_kavinski]

I am going to reinvestigate my 14900K today, give it a second chance.

Keep this in mind too: https://skatterbencher.com/2023/01/...3900ks-overclocked-to-6300mhz/#AC_DC_Loadline

By adjusting the DC loadline to account for the expected difference in voltage between the requested VID and the effective voltage, we ensure the CPU power management unit has the correct information to do its power calculations. If we don’t do this, we end up with power consumption reporting vastly different from reality which may impact the CPU’s Turbo Boost behavior.

 

[Spicedaddy]

Find out the 5.8GHz and 6.0GHz VIDs, that's what shows if it's good or not.

On a 280 AIO, I'd use Intel limits and try to undervolt so it keeps stock frequencies in CB R23 multicore. Single core shouldn't be a problem temp-wise.

 

[Rigg]

My 14900K came in today. What a piece of crap. 1.438VID for 5700MHz. Even set to 175 Watts P0 is hitting 100C with my 280 AIO with HT off! I'll check thermal contact but I'm sure this particular sample is a dog. It's going right back to Amazon without delay.

You should order a new motherboard while you're at it. The VRM on your motherboard is inadequate for a 14900k. The VID is adjustable in the BIOS and not necessarily reflective of how good the chip actually is.