Discussion Raptor Lake Build Thread

[A///]

dookie are those trident zeros or ripjaws?

 

[Carfax83]

@Carfax83 , anything about like 2000 or 3000 runs in gear mode 2 instead of 1:1, right ? so you are at that ? . Using DDR4, on my 12700F, I can not go over 3000 I think. Or does DDR5 eliminate that problem ?

With DDR5 on Alder Lake or Raptor Lake, you can't run at 1:1 ratios at all. The CPU automatically defaults to 1:2 and as far as I know, it can't be changed.

So all of my scores at DDR5 7400 have the memory controller running at only 1850. If the memory controller was running at 2000, that would mean the memory itself would be at 4ghzx2 = 8000mhz. 3000 in gear 2 would mean the memory would be running at 12ghz.

 

[Carfax83]

I'm running my 7200 at 6800 CL32 currently. Still trying to get higher speeds to run. Latency is at 64.9ns. I have to find the right settings to get it to run higher on Z690.

Think I'll tweak that tRAS myself to see if I can get latency down.

I think it works, but you have to decrease it significantly. So far I've done tRAS 100 and it did help with my latency score. Going to try 90 and see if I can get into the 5xns range.

 

[Hulk]

I also am cooling a 13900K with a U12A. It can keep up with the stock volted 13900K in my case to about 180W before throttling. Except for Handbrake I'm not seeing any of my apps hitting this level of compute load so outside of benchmarks I don't think more cooling would matter much in my day-to-day work.

 

[Carfax83]

Maybe revert the CPU back to the default settings and run the test again? It's possible that the undervolt and power limitation is somehow slowing down the IMC.

It could very well be the power limits I am putting on the CPU. But @DooKey had a similar latency as mine with DDR5 6800 CL32, and I don't know if he was undervolted or underclocked.

Another thing is the version number. Aida64 explicitly says that you can't really compare between version numbers because the changes in the code between versions can make significant differences as to the score.

That said, I think my experiment didn't really bear much fruit. tRAS doesn't affect latency as much as I thought it did, although it definitely shaved off a few ns, and to get into the sub 60ns range I would probably have to monkey with the other timings and I honestly don't care enough to try.

One thing I noticed is that my bank cycle time dropped by 25 clocks, according to CPUID. At the stock tRAS, bank cycle time was 155 and now it's 130.

 

[DooKey]

dookie are those trident zeros or ripjaws?

Trident Z5s

 

[DrMrLordX]

Extended loads on the system are always going to create heat soak if you're not exhausting the air and cycling in fresh cooled air frequently.

Absolutely. But you also have to consider the entire system and look for weak points. With a WC setup your main limitations are the IHS (if you don't delid) and your block. You can raise flow rate to the point that your coolant is barely a limitation, and you can always throw more rad and/or more airflow at the problem if you're willing to mount your rad(s) externally.

With an HSF it's the IHS, base, heatpipes, and fins. You can shove more air through the fins, making the limitation unique to that system your heatpipes. And you can't just add more pipes, you get whatever you get from the manufacturer. Plus there are physical limitations restricting how many heatpipes can realistically get good access to the heat flux coming through the IHS to the base.

Taken as a whole, WC setups have much better potential to cool since you're replacing the base + heatpipes with a block. Those blocks work very well since they force heat to travel a much shorter distance before reaching coolant (in this case, water).

isn't 11th gen outclassed by 13th gen in both gaming and productivity by a big margin?

Yes, definitely. An 11900k is barely worth consideration.

 

[Carfax83]

I'm running my 7200 at 6800 CL32 currently. Still trying to get higher speeds to run. Latency is at 64.9ns. I have to find the right settings to get it to run higher on Z690.

Think I'll tweak that tRAS myself to see if I can get latency down.

Have you considered switching over to a Z790 motherboard? The Z790 motherboards handle high speed memory far better than the Z690 ones as the latter has a tendency to struggle. Case in point, this Hynix A die memory is just absurdly easy to overclock. Stock value is DDR5 7200, but I have it now at DDR5 7800!!

A 600mhz overclock without touching the voltage. Just adjusted the primary timings and that was it.

 

[igor_kavinski]

Stock value is DDR5 7200, but I have it now at DDR5 7800!!

AIDA64 screenshot or it didn't happen

 

[DooKey]

Have you considered switching over to a Z790 motherboard? The Z790 motherboards handle high speed memory far better than the Z690 ones as the latter has a tendency to struggle. Case in point, this Hynix A die memory is just absurdly easy to overclock. Stock value is DDR5 7200, but I have it now at DDR5 7800!!

A 600mhz overclock without touching the voltage. Just adjusted the primary timings and that was it.

I thought about switching over, but I don't believe the gains would be worth the effort or the cost. A few more frames when I already get more than enough can wait until the next round of upgrades. I'm pretty happy with the rig as it sits.

 

[Carfax83]

I have to say, I am very pleased with how my new rig turned out. It was everything I wanted it to be in terms of performance, how cool it runs and efficiency and is a literal quantum leap above my previous rig. Alder Lake and Raptor Lake have a bad reputation for being power hogs, but the both of them can be easily tamed if you're willing to undervolt and or underclock. Having done both, my CPU is still blazing fast at 5.2ghz for the P cores and 4.3ghz for the E cores, yet the package power is just 180w at less than 1.2v when maxed (see my HWinfo screenshot on the previous page) and the package temp is 84c with air cooling. I call that efficient for a CPU with this many cores running at 5.2ghz. The memory controller is also top notch, offering great performance and stability running high frequency DDR5.

So all in all, I highly recommend Raptor Lake to anyone

 

[Markfw]

I have to say, I am very pleased with how my new rig turned out. It was everything I wanted it to be in terms of performance, how cool it runs and efficiency and is a literal quantum leap above my previous rig. Alder Lake and Raptor Lake have a bad reputation for being power hogs, but the both of them can be easily tamed if you're willing to undervolt and or underclock. Having done both, my CPU is still blazing fast at 5.2ghz for the P cores and 4.3ghz for the E cores, yet the package power is just 180w at less than 1.2v when maxed (see my HWinfo screenshot on the previous page) and the package temp is 84c with air cooling. I call that efficient for a CPU with this many cores running at 5.2ghz. The memory controller is also top notch, offering great performance and stability running high frequency DDR5.

So all in all, I highly recommend Raptor Lake to anyone

OK, not trying to rain on your parade, as that sounds much more efficient than stock. But my 7950x's are running 142 watt, and the benchmarks are almost the same as at stock.

My point is, that both Raptor Lake and Zen 4 can run with far less wattage than stock, but Zen 4 appears to have the lead in efficiency still.

 

[Carfax83]

OK, not trying to rain on your parade, as that sounds much more efficient than stock. But my 7950x's are running 142 watt, and the benchmarks are almost the same as at stock.

My point is, that both Raptor Lake and Zen 4 can run with far less wattage than stock, but Zen 4 appears to have the lead in efficiency still.

For the kind of stuff you're doing with your computers Mark, yeah Zen 4 would be more efficient. If I recall correctly, you're into distributed computing right? So very parallel workload where the cores are pegged to the max all the time. That kind of workload favors CPUs with high core counts of the big variety.

Before Zen 4 and Raptor Lake released, I said that Zen 4 would excel in those workloads, because the 5950x already excelled in them. But for gaming and many normal desktop applications, Raptor Lake is going to offer better performance per watt when it's tuned and undervolted. Raptor Lake CPUs still have lower idle power usage and have higher performance per watt in gaming and lightly threaded workloads, which is what I mostly use my PC for.

The only time when my cores will be pegged is when I do some encoding/transcoding, which is infrequent. Also, it's amazing that Raptor Lake could even come close to the 7950x in terms of raw performance in heavily multithreaded apps, but of course to do this it had to blow out the power efficiency.

 

[MarkPost]

I have to say, I am very pleased with how my new rig turned out. It was everything I wanted it to be in terms of performance, how cool it runs and efficiency and is a literal quantum leap above my previous rig. Alder Lake and Raptor Lake have a bad reputation for being power hogs, but the both of them can be easily tamed if you're willing to undervolt and or underclock. Having done both, my CPU is still blazing fast at 5.2ghz for the P cores and 4.3ghz for the E cores, yet the package power is just 180w at less than 1.2v when maxed (see my HWinfo screenshot on the previous page) and the package temp is 84c with air cooling. I call that efficient for a CPU with this many cores running at 5.2ghz. The memory controller is also top notch, offering great performance and stability running high frequency DDR5.

So all in all, I highly recommend Raptor Lake to anyone

What app/apps do you use to check CPU stability?

 

[DooKey]

Can we please keep this thread to Raptor Lake builds only. We don't need to bring comparisons to AMD into it or things start to go off-track.

 

[lopri]

I picked up 13700K + Gigabyte B660 Aorus AX the other day. I haven't gotten to put it together but am looking forward to it. It's been a while I build a high-end Intel system for my main.

Another purchase: Arctic Liquid Freezer II 420.

 

[lopri]

Microcenter has 13900K in stock, btw. For some reason their website is not up to date, but I saw them at the store.

 

[DrMrLordX]

Microcenter has 13900K in stock, btw. For some reason their website is not up to date, but I saw them at the store.

That's more common than you'd think in retail. Someone isn't scanning in stock properly for whatever reason.

 

[Carfax83]

What app/apps do you use to check CPU stability?

I just used Cinebench and a few games plus regular browsing. Cinebench tests CPU stability on at a high power draw, while gaming and browsing at moderate to low power draw. All of them are important, especially because my CPU is undervolted.

Whenever you undervolt a CPU, you can't just rely on Cinebench to test stability. In the past I've used Prime95 and other apps to test stability and while it passed, I got crashes during simple things like idling, browsing and gaming. That's because the CPU was undervolted and the voltage would drop too low during idle and low power states.

So far I haven't had a single CPU related crash yet since I dialed in my settings. During the initial testing phase I did though. I had a memory related crash two days ago while gaming when I tested my RAM at 7800mhz without adjusting any voltages. Switching back to DDR5 7600mhz fixed that problem.

 

[igor_kavinski]

That's because the CPU was undervolted and the voltage would drop too low during idle and low power states.

Is there some setting in BIOS to prevent the voltage drop that you are using now?

 

[Carfax83]

Is there some setting in BIOS to prevent the voltage drop that you are using now?

I think that's what Loadline calibration does (more so for the high power draw though), but I haven't messed with that setting as I haven't needed to.

The only CPU related settings I've changed are:

1) PL1 and PL2 are set to 220w

2) Negative offset for the V core at -0.100v

3) Clock speed on the P cores is limited to 5.2ghz

Raptor Lake has some built in clever power management that seems to manage the power draw and performance very well. That's why I'm so impressed with it. Before I installed it, I thought it would take at least one or two weeks for me to dial in the final settings, but it only took me a day. I based that on my experiences with previous Intel CPUs, especially Sandy Bridge-E, Ivy Bridge-E, Haswell-E and Broadwell-E. I had all of those HEDT CPUs and all of them were a pain to tweak.

The other major part of the equation of course is that Intel's 10+++ or whatever node they are using for Raptor Lake is actually really, REALLY good. I am getting a 32 core CPU with 8 P cores at 5.2ghz and 16 E cores at 4.3ghz at only 180w at less than 1.2v which is pretty damn good.

I know that this thread is supposed to be Raptor Lake builds only and I agree, but I'm curious as to what clockspeeds @Markfw is getting out of his 7950x at 142w.

 

[Markfw]

right now, with all 32 threads@100% load (WCG) its running 4.8 ghz at 142 watts.. (for those who don't know, @230 watts, base is 4.5, and max single core boost is 5.7)

Yes, I will try not to post any Zen4 except when requested.

 

[Carfax83]

right now, with all 32 threads@100% load (WCG) its running 4.8 ghz at 142 watts.. (for those who don't know, @230 watts, base is 4.5, and max single core boost is 5.7)

Yes, I will try not to post any Zen4 except when requested.

Thanks, can you also tell me the voltage the CPU is using?

 

[Markfw]

Thanks, can you also tell me the voltage the CPU is using?

Not sure running linux, how do I tell ? lscpu only gives speed and other info.

Edit : duh... I have another 7950x and its running windows. 1.14 by Ryzen master, also 4.8 ghz, also 32 cores@100%, and its set to 85c.

@Carfax83 I updated this.

 

[Markfw]

@Carfax83 what were your cb23 scores ? Now I am curious. Once you post, I will post in the Zen 4 builders thread, to keep this thread clean.