Trident Z5 DDR5 7200mhz 2x16gb with Ryzen 7600x - timings and benchmark

[Woomack]

Believe me, if there were any good free tools, then the community would use them and flood all forums. Somehow, they're not repeatable or provide confusing results for wider comparisons. Some programs are in use only by competitive overclockers, but there are fewer of them each year.
OCN is a place where it is hard to find any good info. It's not because there isn't, but because most members are totally clueless and only flood forums with random stuff or disagreement without any proper info. There are 500+ post threads where most are totally useless or misleading.

I agree that the last generation of MSI motherboards is not the best. However, it's still better than the last ASRock generation (comparing average, not specific models). I can't say about the BIOS for some motherboards, but those higher series were pretty good. It's the same with other brands. Lower or less popular models get worse support. ASUS or Gigabyte keep the same updates for more models (like all Strix and higher series have the same BIOS releases and updates), but other brands provide better support only for top models. Like ASRock focuses mainly on Taichi, MSI on the MEG series, and EVGA on Dark/Dark KP ... Biostar doesn't focus on anything, as they release one BIOS, and good luck.
ASRock motherboards right now are fine. Not great, but good for most users. I have no idea if Nick Shih (he was their in-house overclocker and was leading the BIOS team) moved somewhere else, but they have no OC series motherboards, no OC improvements, and no additional OC options in BIOS in the last two generations. AMD motherboards still overclock RAM well, as you can push them to 8000, the same as on all higher competitive motherboards. However, Intel motherboards are pretty bad in memory OC, and training.

ASRock B650E PG ITX works well, but it has very slow training, very slow boot time, terrible coil whine in idle, and totally stupid 12k+ RPM fan on the M.2 radiator, which ASRock advertised as some magical technology.

Btw. I left the PC running at work yesterday and forgot about it. It passed a 3h+ AIDA64 stability test at 8000 CL36-48-48 1.45V. I will play with sub-timings when I find some more time.

 

[delvechio212]

Tbh ive been using ASUS and Asrock boards in my previous builds and for some testing, and i always found they managed to go over their listed specs. Good vrm thermals, good oc support and various features in their bioses.
For ZEN4 and DDR5 ASUS has always had GDM on off support, just a retarded name for it.
Its under ram timings, way under all the timings and you have AUTO-BUFF-UNBUFF options , buffered is on, unbuff is off for gdm (or other way) so i was using it since i got this board on it stock and updated bioses.
Also newest bios and agesa update did drastic changes in stability and overclock possibilities for ram on asus boards.

Also for precision boost , leave it at auto in bios and use hydra from 1usmus

 

[Woomack]

Until the latest generation, most motherboards could make much higher RAM frequencies than specified. Since the DDR5 release, it looks like maybe in 70%, the max memory clock in specs matches the max stable clock on at least an average CPU. Every ASRock or MSI motherboard tested by me could make as much as in specs or, in the best case, one step above. Like the Z690 Unify-X has 6800 max, it was working at 6800 on one slot and 7200 on another one. ASRock Z790 Steel Legend had 7200, and it couldn't even boot above that. Gigabyte Z790 had 7600 in specs just after release (multiple mobos), and all could reach 7800-8000. These are just examples as the list is much longer. I was really disappointed by everything Z690 while Z790, even cheaper, is pretty good. Many B760 go up to 8000 too.

All AMD motherboards (count the leading brands) had a max of 6200-6600 (good luck getting a CPU that can make 6600), and all could make 6200-6600. There were different results in synthetic tests as some were setting relaxed sub-timings, but in general, the frequency was about the same since release (well, since stable BIOS releases, as some first releases were pretty bad). Surprisingly, after unlocking higher clocks, most even cheaper motherboards still can go up to 7800-8000, and barely any can pass 8000. Differences are mainly in RAM training and sub-timings. Some mobos are acting weird or are missing some options in BIOS. You can always check if the AMD OC tab has something that you are missing, but sometimes it's hard to find it.

I found a promo on 7800X3D, and I ordered it today. I have no idea why but all Ryzens went up by 10% in the last 3-4 days and were slightly going up for nearly 2 weeks. For some reason, one store lowered the price only for 7800X3D while other AMD CPUs went up by 10%+. I didn't want to buy any CPU, but I was talking with my brother and I will upgrade his PC a bit, so I have a chance to replace something for tests.

 

[Woomack]

ASRock B650E PG ITX has a GDM option, but the motherboard doesn't even start when it's disabled at more than 7200. At 7200, it doesn't change anything.
The DDR5 Nitro option is weird. In theory, higher values should help with frequency at the cost of latency. In real when it's disabled, then 8000 isn't stable. When I play with these settings I see no difference at all. Latency is the same, and I can only lose stability at 8000.

On Ryzen 7800X3D, I see worse latency in general, and it's 1 CCD CPU, so memory read and copy are about 30GB/s worse than on my 7950X. Latency is about a 10ns worse on the 7800X3D than on the 7950X. At first, I thought that a 2x48GB Micron kit was so bad as I had 93ns at EXPO and 77ns after tweaking. When I switched to 2x24GB Dominators, I saw about 10ns worse latency than I had before.
I'm a bit surprised as 7800X3D has memory bandwidth like 7600 (1CCD), but 7600 has latency like 7950X. 7950X has bandwidth like 7600X (or even worse) but better latency.

 

[delvechio212]

Hi everyone,

I have some new news with the Trident Z 7200 2x16gb kit and 7950x this time.

I updated bios to PRIME X670-P BIOS 1811 and i was getting some weird errors, bsod-s and random hickups, bullshit i couldnt even describe.. Hangs and whatnot.
I reverted back to PRIME X670-P BIOS 1808 , which is one version newer from what i had.

Here are the 7000mhz timings on 7950x, stock frequencies on cpu just applied -10 curve to it, nothing more.

I also noticed bios has a new preset for 7600mhz micron 2x16 and 7600mhz micron 2x24 ram timings which are found on those two new versions.
I could boot at 7600,7200, 7000 with those presets but got some weird hangs, random bsod, low scores in any gpu benchmarks so i went with my own timing configuration.

Will keep you all updated on what i can achieve with this new cpu and its timings.

Btw i have two 7950x with me, one can do 2200 FCLK without any problems, second one can do only 2100 but has better cppc in hydra.

 

[Woomack]

I also noticed bios has a new preset for 7600mhz micron 2x16 and 7600mhz micron 2x24 ram timings which are found on those two new versions.
I could boot at 7600,7200, 7000 with those presets but got some weird hangs, random bsod, low scores in any gpu benchmarks so i went with my own timing configuration.

This is interesting. I haven't seen any new profiles on my mobos. I wonder if it's a new IC or if they improved something. The 2x48GB Micron kit that I recently tested could reach 6800 but at very relaxed timings. 2x16GB/24GB can definitely make more.

Btw i have two 7950x with me, one can do 2200 FCLK without any problems, second one can do only 2100 but has better cppc in hydra.

My 7950X could do 2100 too. 7800X3D can make 2200. Both 7600 can make 2130 or 2160 ( I don't remember exactly, but a bit above 2100).

 

[delvechio212]

Wow 8000mhz with fclk 2000 achieved on bios 1811 on asus x670-p board.

Looks like its bechmark time.. will get back to you with scores and more stability tests

 

[Woomack]

8000 seems to work pretty well on many motherboards that are not designed for that. Problems are with anything higher, as most motherboards won't even start at 8200. I had no MSI AM5 mobo, but ASRock, ASUS, and Gigabyte all run at 8000. It's a bit weird that the 1 memory ratio makes such a big difference as full stability -> no boot at all. I wonder if new APUs support higher memory like it was in AM4.

 

[delvechio212]

8000 seems to work pretty well on many motherboards that are not designed for that. Problems are with anything higher, as most motherboards won't even start at 8200. I had no MSI AM5 mobo, but ASRock, ASUS, and Gigabyte all run at 8000. It's a bit weird that the 1 memory ratio makes such a big difference as full stability -> no boot at all. I wonder if new APUs support higher memory like it was in AM4.

Also i have a support ticket open with ASUS about the 1808 and 1811 bios versions.
Under dram timings there are several presets that board offers, like hynix 7600 2x16 SR 1.4v, hynix 6400 2x16 etc,
if you activate any of the presets, VDD AND VDDQ voltages are set to 1.12V.. EVEN if you change them manually they stay at 1.12V and you cannot do anything to change them.. EVEN if you set it to 1.4v manually bios boots with 1.12 which causes stability issues.

Only way to revert is to clear bios via clr cmos pins, and than manually set all values you want without activating profiles...

ZENtimings window shows voltage as wrong, as well as AIDA64.. Seems to be some bios issue.

 

[Woomack]

ASUS presets don't work for me at all in this generation. In the previous gens, all higher presets were made on cherry-picked CPUs and RAM and I always had to lower one ratio down to make it work.
On Crosshair X670E Gene, I couldn't make any profile to work. The same Maximus Tweak on Z790 is acting weird. I don't know why, but profiles simply don't work right on all AM5 and Z790 motherboards that I tested. On the other hand, Gigabyte and MSI profiles don't work too MSI Z590 was great and profiles were saving a lot of time, but Z690 and Z790 are meh.

I think Zen Timings has it even in the FAQ that voltages don't work or may be wrong. Some other things can be wrong too. The last version is pretty old if I'm right. AIDA64 may read from the wrong sensor. It sometimes happens. Hwinfo64 is usually the best for sensor reading.

I noticed that not many timings actually affect AMD's performance. The same single/dual rank doesn't change much. I'm glad that the 1:2 ratio can be even faster if set at a high enough memory clock as it gives something more to play with than a 6400 1:1, and also gives hope on something even better when new CPUs are released. AM5 and current motherboards are supposed to last some more, and I hate to switch motherboards only because one little feature was added.