Share your AIDA 64 cache and memory benchmark here

[vintologi]

Adjusted the primaries, bumped up the clock, but need a little more insight into secondary and tertiary timings.

View attachment 344651

You managed to beat my read score but otherwise i still score better at 6600 CL30.

You almost certainly cannot even boot if you try to copy my timings since a lot of my timings are pretty tight (it's for 2x32 A-die).

Try setting tREFI to 65528 first. Should give you a significant boost easily.

Also: run some stresstests to verify it's at least decently stable.

Reduce your timings to the XMP default and see if it still works, i could set almost all to XMP or lower despite 10% frequency overclock.

I think there is something wrong with my setup on my end too, high latency and poor read performance, only 15% better than XMP @ 6000.

 

[dgianstefani]

TBH if you're on a Raptor Lake K chip i7/i9 and have a 1DPC Z790 with A die, you should be going for at least 7200, preferably 7600+ or even 8000+.

6600 MT is Z690/12900K numbers, not terrible, and still better than Zen, but very conservative compared to what the platform can do.

$2-400 ITX boards beating $500 2DPC ATX boards will never not be funny to me. A free OC wet dream memory trace layout that you otherwise have to pay extra for with the 1DPC ATX boards lol.

 

[RAZOR1911]

1,57 ist nicht hoch für B-Die.

Sie können bis zu 1,7 V täglich sicher betreiben, wenn es kühl gehalten wird.

But he did something to the chipset and CPU voltages, so I ask what they do on a daily basis. and since I have SK Hynix on my modules: G Skill TridentZ RGB F4-4000C18-32GTZR, I wanted to ask what I should set up

 

[freeagent]

Aber er hat etwas mit dem Chipsatz und den CPU-Spannungen gemacht, also frage ich, was sie täglich machen. und da ich SK Hynix auf meinen Modulen habe: G Skill TridentZ RGB F4-4000C18-32GTZR, wollte ich fragen, was ich einrichten soll

We speak English here, please use a translator

 

[RAZOR1911]

But he did something to the chipset and CPU voltages, so I ask what they do on a daily basis. and since I have SK Hynix on my modules: G Skill TridentZ RGB F4-4000C18-32GTZR, I wanted to ask what I should set up

 

[vintologi]

TBH if you're on a Raptor Lake K chip i7/i9 and have a 1DPC Z790 with A die, you should be going for at least 7200, preferably 7600+ or even 8000+.

6600 MT is Z690/12900K numbers, not terrible, and still better than Zen, but very conservative compared to what the platform can do.

That's true for single rank (2x16 or 2x24) but 2x32 is a lot harder to overclock so just getting to 6600 can be a challange on a lot of motherboards (i couldn't do it until i updated bios to A92).

You can go maybe 12% higher in frequency with single rank on good motherboards.

The highest i have seen decently stable with dual rank is 7400 CL34

$2-400 ITX boards beating $500 2DPC ATX boards will never not be funny to me. A free OC wet dream memory trace layout that you otherwise have to pay extra for with the 1DPC ATX boards lol.

Doesn't make sense to pay 500$ for a motherboard when the platform itself is a dead end (with gen 14 being the last).

Connectivity is limited regardless due to the limited connectivity the CPU itself is offering.

 

[Arctucas]

You managed to beat my read score but otherwise i still score better at 6600 CL30.

You almost certainly cannot even boot if you try to copy my timings since a lot of my timings are pretty tight (it's for 2x32 A-die).

Try setting tREFI to 65528 first. Should give you a significant boost easily.

Also: run some stresstests to verify it's at least decently stable.

Reduce your timings to the XMP default and see if it still works, i could set almost all to XMP or lower despite 10% frequency overclock.

I think there is something wrong with my setup on my end too, high latency and poor read performance, only 15% better than XMP @ 6000.

View attachment 344714

A little more tuning:

 

[vintologi]

A little more tuning:

View attachment 344836

That's much better.

Latency is ok but you want lower of course. Aida64 on windows is not the most reliable benchmark, varies significantly between every time you run it.

 

[dgianstefani]

That's much better.

Latency is ok but you want lower of course. Aida64 on windows is not the most reliable benchmark, varies significantly between every time you run it.

Depends how bloated your OS is.

I get a consistent 0.1 ns difference between runs, and only 0.4 ns between safe mode and normal boot.

 

[vintologi]

Depends how bloated your OS is.

I get a consistent 0.1 ns difference between runs, and only 0.4 ns between safe mode and normal boot.

I guess i need to debloat windows 10 more. For me it's like ±0.6 ns.

I do get 10ns lower latency with intel mlc when running linux.

I also look into disabling core parking and other attempts to improve performance. I use linux most of the time so i have not really optimized windows for the best performance.

 

[Tatili]

1.57 isn't high for B die.

You can run safely up to 1.7 V daily if it's kept cool.

Hi guys, precisely kept cool indeed

 

[dgianstefani]

I guess i need to debloat windows 10 more. For me it's like ±0.6 ns.

I do get 10ns lower latency with intel mlc when running linux.

I also look into disabling core parking and other attempts to improve performance. I use linux most of the time so i have not really optimized windows for the best performance.

I use Windows 10 IoT Enterprise LTSC, which starts off pretty clean compared to normal consumer 10 or 11.

Calypto's latency guide helps, as does other miscellaneous tuning and BIOS stuff, but you don't need to do some of the mor extreme things like disabling AV.

Calypto's Latency Guide

docs.google.com

 

[Tatili]

But he did something to the chipset and CPU voltages

Hi enthusiast, YES call "something" named means BCLK OC and it's only a route for me to increase in the same time, cores speed & mem freq with X3D on my budget B550 Hawk mobo

so I ask what they do on a daily basis.

I did posted couple pictures in test, attempting to Wheas free at IF2000, which are almost impossible to get rid of them for X3D at 4000
I already did it at 3952mhz by BCLK, so here is my actual daily & gaming 24/7 Profile

and since I have SK Hynix on my modules: G Skill TridentZ RGB F4-4000C18-32GTZR, I wanted to ask what I should set up

And NO sorry can't help you with CJR dies, more info should being already posted by our Lord Micron & CJR DDR OC @The_King here

 

[dgianstefani]

Hi enthusiast, YES call "something" named means BCLK OC and it's only way to increase cores speed & mem freq in same time on X3D


I did posted couple pictures in test, attempting to Wheas free at IF2000, which are almost impossible to get rid of them for X3D at 4000
I already did it at 3952mhz by BCLK, so here is my actual daily & gaming 24/7 Profile


And NO sorry can't help you with CJR dies, more info should being already posted by our Lord Micron & CJR DDR OC @The_King here

Some boards have an external BCLK generator so you can OC the CPU separately from everything else like PCIe bus/memory etc. Annoying that they're generally not in ITX.

 

[Arctucas]

Trying to tighten timings.

One thing I have noticed, many users list low tRAS values, but when I try to lower tRAS on my rig and run test, system reboots.

I have seen two formulas to determine tRAS;

tCL + tRCD + tRTP = tRAS

or

tRCD + tRTP = tRAS.

The first works for me, but the second (lower value) does not.

 

[opheen]

7600c34

 

[Arctucas]

Apparently, I had a corrupted Windows install. Started having boot issues and crashes.

Reset the OS and now I can set tRAS to tCL + tRTP.

Weird.

Any tips on improving?

 

[dgianstefani]

Apparently, I had a corrupted Windows install. Started having boot issues and crashes.

Reset the OS and now I can set tRAS to tCL + tRTP.

Weird.

Any tips on improving?

View attachment 345104

Use Karhu for stability tests.

Unstable memory will corrupt Windows.

Other memory software is good, but won't always find issues when they exist. Karhu stable is 100% stable, unless conditions such as temperature changes between daily use and RAM OC sessions.

 

[Arctucas]

Use Karhu for stability tests.

Unstable memory will corrupt Windows.

Other memory software is good, but won't always find issues when they exist. Karhu stable is 100% stable, unless conditions such as temperature changes between daily use and RAM OC sessions.

OK, bought a Kahru license, will try it out.

Anyway, do you see any settings you might suggest I adjust to help improve performance?

Had to go back to 7600.

 

[mrcardio33x]

6150 MT/s CL26 35 32 30 / 2221 FCLK via BCLK 102.5
VDD: 1.55v VDDQ/IO: 1.35v VSOC: 1.23v
Nitro 1 2 0

CPU: BCLK 105 - LLC7 - PBO Scaler X6 - Boost Clock (-) 150 MHz or 5150 MHz

* Note Zen Timings doesn't show correct 102.5 BCLK OC.

 

[freeagent]

Took the truck out for a rip

 

[Braegnok]

BIOS 1905 AGESA 1.1.0.2b testing 8000MTs CL34 profile.

 

[RayneYoruka]

Hello
My values are like this with active GDM (Gear Down) the question is, is a newer AGESA worth more when writing?

I can't turn off GDM because windows crashes.


View attachment 343201

View attachment 343203


Was that stable? are these also SK hynix chips? and what tension did you have there? is this permanent or just a benchmark?

I'm genuinely surprised that you're getting almos the same numbers as I do with the gskill trident Z neo at 3600mhz vs 4000, this always amazes me

Here you can see my post with my 5900x and previously with my 3700x

Share your AIDA 64 cache and memory benchmark here

Just messin around..

www.techpowerup.com

 

[Atmaniac]

Other memory software is good, but won't always find issues when they exist. Karhu stable is 100% stable, unless conditions such as temperature changes between daily use and RAM OC sessions.

I prefer the VT3 Vector Transform stress test in Y-Cruncher for a RAM stability check. It's the hardest test I've seen to date for that purpose.

 

[Dia01]

BIOS 1905 AGESA 1.1.0.2b testing 8000MTs CL34 profile.

View attachment 345536 View attachment 345537

Nice