Share your AIDA 64 cache and memory benchmark here

[mrcardio33x]

VDD: 1.55v

 

[InVasMani]

Is there much consensus yet on what's generally deemed safe voltages for memory IMC/VDD/VDDQ/TX for DDR5 it seems like 1.55v VDD is a bit high for long term usage. I haven't used more than 1.45v on any of those. I'm just curious what most would deem generally safe for long term usage on them with DDR5. I get that people can't definitively say will absolute certainty, but at the same time there is probably some consensus by now what has been work well in general for people and a maybe a few fringe isolated incidents of a particular voltage or voltages being too aggressively pushed leading to issues.

 

[mrcardio33x]

Is there much consensus yet on what's generally deemed safe voltages for memory IMC/VDD/VDDQ/TX for DDR5 it seems like 1.55v VDD is a bit high for long term usage. I haven't used more than 1.45v on any of those. I'm just curious what most would deem generally safe for long term usage on them with DDR5. I get that people can't definitively say will absolute certainty, but at the same time there is probably some consensus by now what has been work well in general for people and a maybe a few fringe isolated incidents of a particular voltage or voltages being too aggressively pushed leading to issues.

VDD @ 1.55v isn't even remotely "too high"
There are even EXPO kits at 1.55v for VDD

VDD is only "too high" if you can't keep temps down.

I have a Noctua 2200 RPM 80mm fan over my DIMMS running @ 1.55v and max temps are ~ 43c under load (gaming w/ RTX 4090 at 450w+ and or stress test such as Karhu, TM5, etc.)

 

[Det0x]

Is there much consensus yet on what's generally deemed safe voltages for memory IMC/VDD/VDDQ/TX for DDR5 it seems like 1.55v VDD is a bit high for long term usage. I haven't used more than 1.45v on any of those. I'm just curious what most would deem generally safe for long term usage on them with DDR5. I get that people can't definitively say will absolute certainty, but at the same time there is probably some consensus by now what has been work well in general for people and a maybe a few fringe isolated incidents of a particular voltage or voltages being too aggressively pushed leading to issues.

2.07 VDD @ CL30 8000MT/s is too high

But i have been running 1.74/1.6 daily for almost 2 years now, and no issues with that

 

[freeagent]

^^

What a beast! Nice.

Still kickin it old school here..

 

[InVasMani]

2.07 VDD @ CL30 8000MT/s is too high
View attachment 359265

But i have been running 1.74/1.6 daily for almost 2 years now, and no issues with that

Seems I was being over conservative on DDR5 voltages. What's temps are advised to keep DDR5 below!?

 

[mrcardio33x]

Seems I was being over conservative on DDR5 voltages. What's temps are advised to keep DDR5 below!?

Yeah, as long as you have sufficient cooling I wouldn't be too concerned about VDD voltage. Plenty of us have been running 1.55v + VDD for over a year now with no issues whatsoever.

Worse case scenario is you out ~ $100 for new RAM regardless so not exactly the end of the World. Highly unlikely however.

 

[Solaris17]

 

[Sabotaged_Enigma]

View attachment 359265

This cache speed... Is it manually overclocked to run all cores at 5145 MHz? Or is the ES version even better than retail ones?

 

[Det0x]

This cache speed... Is it manually overclocked to run all cores at 5145 MHz? Or is the ES version even better than retail ones?

All Zen5 CPUS have 2x L1 cache bandwidth compared to Zen4

 

[Sabotaged_Enigma]

All Zen5 CPUS have 2x L1 cache bandwidth compared to Zen4

I know. But that's still higher than a default 9950X.

 

[mrcardio33x]

New PR!

VDD: 1.55v

 

[AVATARAT]

 

[sethmatrix7]

Trying to see what I can get with 1.2vsoc and 1.35VDD. Passed 3 hours of OCCT and a dozen y-cruncher pi runs.

Advice welcome...

 

[mrcardio33x]

Safe Mode vs. Windows 11 Pro - Insider Program - Release Preview - 24H2 Version.

 

[sethmatrix7]

Safe Mode vs. Windows 11 Pro - Insider Program - Release Preview - 24H2 Version.

View attachment 361240View attachment 361241
View attachment 361242View attachment 361243View attachment 361244

Does lower tRAS or tRC cause instability for you?

 

[lolek86]

View attachment 347897

Decided to make cool, stable, latency oriented profile so 6000CL26 and achieved personal record of 54ns latency. Amazing how tight timings you can get at 6000 with only 1.53 vdd all else voltages almost stock and I still could try few things in secondary timings. This is fully stable and without any direct Ram cooling. Next step how much of this conditions could be replicated at 6200.

 

[natr0n]

trying out some ideas and being successful. This imc can handle anything.

 

[mrcardio33x]

Decided to make cool, stable, latency oriented profile so 6000CL26 and achieved personal record of 54ns latency. Amazing how tight timings you can get at 6000 with only 1.53 vdd all else voltages almost stock and I still could try few things in secondary timings. This is fully stable and without any direct Ram cooling. Next step how much of this conditions could be replicated at 6200.

View attachment 361642

Not bad,

The non-X3D CPU's hit lower latency due to higher boost clocks.
The R9 7900X I had previously consistently hit Aida64 latency results ~ 4ns lower running the same RAM / Timings vs. my current R7 7800X3D.

So my R9 7900X would be ~ 51ns running my current setup.

 

[lolek86]

Not bad,

The non-X3D CPU's hit lower latency due to higher boost clocks.
The R9 7900X I had previously consistently hit Aida64 latency results ~ 4ns lower running the same RAM / Timings vs. my current R7 7800X3D.

So my R9 7900X would be ~ 51ns running my current setup.

Yeah cpu frequency has direct influence for aida latency. If i would run same memory setup with stock cpu at 5150 instead of current 5500 , aida would show about 3ns worse results.

I see you run vsoc at 1.24 and fclk at 2200 for 6200 profile above. Did you try lower vsoc value or is necessity? My imc is getting picky if I try 2200fclk with something above 1.2 vsoc.

 

[mrcardio33x]

Yeah cpu frequency has direct influence for aida latency. If i would run same memory setup with stock cpu at 5150 instead of current 5500 , aida would show about 3ns worse results.

I see you run vsoc at 1.24 and fclk at 2200 for 6200 profile above. Did you try lower vsoc value or is necessity? My imc is getting picky if I try 2200fclk with something above 1.2 vsoc.

1.240 VSOC is lowest stable for me @ 6200 MT/s.
At 6000 MT/S 1.185 VSOC works.
At 6400 MT/s 1.28 VSOC is required but seeing as I can't run GDM disabled nor Nitro 1 2 0 here, 6200 MT/S is better in my case.

My previous R9 7900X would boost to 5700 MHz vs. 5150 HMz for my R7 7800X3D
Without the ( - ) 150 MHz Boost Clock Limit my R7 7800X3D would boost to 5300 MHz but would not to 100% stable here, would crash SHA3 for example.

At 5300'MHz I'd likely hit ~ 55.0ns but I prefer to run a 100% stable CPU.

* Previously showed 1.1185 VSOC for 6000 MT/s - corrected to 1.185 VSOC.

 

[Det0x]

After pulling my hair for nearly two weeks i finally feel confident with my 8200MT/s tune and its ready to be shown in primetime
(in the end it turned out my too low tRAS + tRC values was the reason for my very late karhu fails)

Hardware used:

• 9950X Zen5 ES
• ROG CROSSHAIR X670E GENE @ public bios 2303
• 2x16GB Hynix A-die -> GSKILL Royal 8000MT/s EXPO @ 8200MT/s CL32-46-40-48-88 at 1.8VDD (need better binned memory)
• 2200mhz FCLK
• Large custom watercooling (moh-ra on a different floor)

The following stresstest complated in a single run:

• 45k Karhu @ ~355mb/s (~9 hours)
• 1 hour y-cruncher 8.5 all mem-tests enabled (daily PBO CO tune -> upto 273w PPT)
• 41 cycles TestMEM 5 1usmus CFG (2.5 hours)
• And to top it of, 1 hour runmemtestpro

Full screenshot with all stresstests completed

 

[boomheadshot8]

holy shiiiiiiiiiiiiit
new ryzen are neat for high frequencies ?

 

[Det0x]

holy shiiiiiiiiiiiiit
new ryzen are neat for high frequencies ?

Even higher memspeed is easier on hynix m-die with GDM enabled, but as always on the AM5 platform, the performance is alittle lower than a-die since more lax timings

Here are 8400MT/s stable

~15hours Karhu @ 340mb/s
1 hour Y-cruncher 0.8.5 all tests (which is really hard to run for Zen5, upto 336w usage)
And to end it off i ran 1.5hours TestMem 5 1usmus cfg

 

[boomheadshot8]

watercooling ? other temps are fine compared to my 10900kf looool (90°C at 280w)