New DDR5 6600 issues

S

-snoopi-

Limp Gawd
Joined
Jul 26, 2022
Messages
196
Just upgraded from
Crucial DDR5 4800 C40 to
TridentZ DDR5 6600 C34

XMP1 works, XMP2 crashes very frequently.
But I have lower benchmark scores in CyberPunk 2077 4K max, with my new RAM even though it should be better as I went from 4800 C40 to 6600 C34.

Any reason why I experienced less performance?

My full specs:
Evga SuperNova P2 1000W
Asus Z690 STRIX-E
12900K stock no OC
Asus Tuf OC 4090 +1000mem +100core
2x16GB TridentZ 6600 C34 XMP1 active
Samsung 980 Pro

Do I need to OC the CPU to have a working XMP2 for my setup? Or should it work regardless and even make the difference? Should I enter RAM OC settings mnaually for more stability or it doesnt matter?

All BIOS, firmware etc is updated.
 
-snoopi- said:
Just upgraded from
Crucial DDR5 4800 C40 to
TridentZ DDR5 6600 C34

XMP1 works, XMP2 crashes very frequently.
But I have lower benchmark scores in CyberPunk 2077 4K max, with my new RAM even though it should be better as I went from 4800 C40 to 6600 C34.

Any reason why I experienced less performance?

My full specs:
Evga SuperNova P2 1000W
Asus Z690 STRIX-E
12900K stock no OC
Asus Tuf OC 4090 +1000mem +100core
2x16GB TridentZ 6600 C34 XMP1 active
Samsung 980 Pro

Do I need to OC the CPU to have a working XMP2 for my setup? Or should it work regardless and even make the difference? Should I enter RAM OC settings mnaually for more stability or it doesnt matter?

All BIOS, firmware etc is updated.
Click to expand...
12900K should have better IMC.
Maybe try to set (force) the SA voltage to 1.0v or 1.05v (default probably at 0.9v) because it's unlocked part and to maintain 6600++ oc the SA should be a bit higher than 0.9v.

FYI, I managed to overclock my hynix M-Die (8gb x 2 sticks of Patriot 4800 c40) to 6800 32-42-42 2T.
Spec: intel 12100F (SA locked at 0.9v), asus b660i strix (bios version 1003, Dec 2021).
 

Attachments

  • 12100F 16gb patriot 6800 32-42-42-38 trfc 380 2T vddtq 1.57v vddq 1.52v tight stable.JPG
    12100F 16gb patriot 6800 32-42-42-38 trfc 380 2T vddtq 1.57v vddq 1.52v tight stable.JPG
    297.1 KB · Views: 0
Ok i will try increasing SA voltage later.

But XMP1 works, XMP2 does not. What is the difference?
 
-snoopi- said:
Freq same as XMP1.
But i think XMP2 changes some other timins like secondary timings?
Click to expand...
Ah, then that explains why the XMP2 are not stable due to possibly tighter secondary timings while the XMP1 is having more auto on the secondary.
You may try my previous suggestion to set SA voltage a bit higher.
 
xDiVolatilX said:
Interesting I wonder why the new faster ram performed worse. In for answers lol
Click to expand...
This beats me to.
Kind of annoys me, koz I spent an extra $AU500 to get this RAM in hopes I will get a few % more fps as the youtube video comparisons show of the various RAM speeds between 4800 - 6600, but it did the total opposite. Very frustrating.
 
xDiVolatilX said:
Interesting I wonder why the new faster ram performed worse. In for answers lol
Click to expand...
There are so much factor, for ddr5:
1. If the cpu & mobo combo still the same, ram changed to faster freq and tighter timings, it could be:
- imc limitation on the CPU (for K series can be stabilized more by messing the SA or IMC voltage
- ram slot (even nowadays, ram slot can be binned haha) on the mobo
- bios capabilities
2. If the cpu & mobo combo change, or getting better, then:
- the ram probably has locked PMIC or the PMIC is of low quality.
- the ram chip is using old revision like hynix M-Die (actually clocked a bit lower than A-Die but the timings can be tighter), or using samsung / micron
 
So I was going to try upping the system agent voltage as mentioned here, but just realised my SA voltage alreay looks higher than whats suggested, mine seems to be on 1.233v, it was suggested here to try 1.05v.

But I never changed my SA before, so is the default 1.223v thats its on too high?
Here is a screenshot:
20221118-113504.jpg
 
-snoopi- said:
So I was going to try upping the system agent voltage as mentioned here, but just realised my SA voltage alreay looks higher than whats suggested, mine seems to be on 1.233v, it was suggested here to try 1.05v.

But I never changed my SA before, so is the default 1.223v thats its on too high?
Here is a screenshot:
View attachment 527596
Click to expand...
I just asked my friend who is asus overclocker team, and he said that "the 1.233v for SA auto is normal, but for that freq (6600mhz), 1.0v is enough to be honest".

But I wonder that auto at CPU Input Voltage is really 1.777v???
 
Ok so if my SA voltage is already much higher than minimum required of 1.0v for 6600 DDR5, does that mean I could have reached the limit of these sticks? Currently it seems I cannot go above 6600 even on the working XMP1 profile. I tried 6800, and it failed nearly instantly in Windows.
 
-snoopi- said:
Ok so if my SA voltage is already much higher than minimum required of 1.0v for 6600 DDR5, does that mean I could have reached the limit of these sticks? Currently it seems I cannot go above 6600 even on the working XMP1 profile. I tried 6800, and it failed nearly instantly in Windows.
Click to expand...
can you show me the SPD tab of CPU-Z or even better, the thaiphoon burner?
if it's Samsung, it's probably reaching it's limit.

if it's Hynix (and probably Hynix M-Die based on you xmp), for 6800 you can try adjusting:
1. vddtq 1.53v
2. vddq 1.51v
3. imc voltage 1.38v
4. SA 0.9v (locked due to i3 12100F)

I used this voltage for fury beast xmp 6000 c40 (chip hynix M-Die)
 
-snoopi- said:
Ok so if my SA voltage is already much higher than minimum required of 1.0v for 6600 DDR5, does that mean I could have reached the limit of these sticks? Currently it seems I cannot go above 6600 even on the working XMP1 profile. I tried 6800, and it failed nearly instantly in Windows.
Click to expand...
Two things

1. do you have the latest bios update?
2. DDR5 has its own voltage controls built into each stick. I suggest you lookup an article on how to tweak voltage for DDR5. Asus probably has their own article about it. I liked this article:
https://www.overclockers.com/ddr5-overclocking-guide/

And after you installed the new RAM, did you clear the CMOS and re-do all of the settings?
 
Last edited:
Yes I 100% updated the Bios firmware.

I did not clear CMOS. Was I suppose to?

Thank you for the link to the guide, I will read that soon.
 
-snoopi- said:
its Hynix, not sure which die though.

View attachment 527660
Click to expand...
Last three digits of number above barcode - if it's "20A" it's A-die, if it's "20M" it's M-die.

You can see this number through the packaging window prior to purchase, these are the modules I used to look for at Microcenter before I moved on to TeamGroup 7200 kits, and plain "green stick" A-die kits. Overall TeamGroup has been superior to G.Skill in terms of OC headroom they leave on a given kit, as well as thoroughness of testing their kits. G.Skill has by many accounts/complaints at OCN, plus my own experience, been extremely tight in not leaving much room for OC beyond rated/XMP speed at least on their DDR5 offerings. If you're happy with XMP/rated-speed, G.Skill is fine.

The newest G.Skill 6400 kits (and above), and new-ish G.Skill 6600 kits (and above) seem to be the threshold where you have a high chance of A-die at the least cost. Finding openbox G.Skill 6600 kits have been some of the cheapest ($200-$225 for 2x16) price-perf A-die kits (short of PITA-to-buy green stick A-die kits from China/Taobao).

1668828257065.png
 
Last edited:
I have the c32 kit z5 at 6000mhz and 64gb dual channel it works good on the z690 and 12700kf on the msi ddr5 edge Wifi.
 
-snoopi- said:
But I have lower benchmark scores in CyberPunk 2077 4K max, with my new RAM even though it should be better as I went from 4800 C40 to 6600 C34.
Any reason why I experienced less performance?
Click to expand...
Cyberpunk 4K maxed will be completely GPU bottlnecked and as such CPU\RAM speed will have no impact on performance and any difference you are seeing will be margin of error\temperature related.
Timespy CPU test show a small to moderate gain from higher RAM speeds but it is quick and fairly consistent run to run unlike most game tests.

If XMP 1 is running the rated speed and stable then great that is all you need to do.
You can compare performance to make sure if you like using aida64 and 3dmark timespy or some CPU heavy games at low res if you prefer but they will need multiple runs averaged.

If you want to get into overclocking RAM for the fun of it and a little more performance in CPU limited sections of gameplay take a look at this.
https://www.overclock.net/threads/o...d-24-7-daily-memory-stability-thread.1794772/
And although aimed at DDR4 this is a great RAM OC guide.
https://github.com/integralfx/MemTestHelper/blob/oc-guide/DDR4 OC Guide.md
 
You must log in or register to reply here.
Back
Top