![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
- Joined
- Nov 11, 2012
- Messages
- 1,752
I had a few 8GB SK Hynix HMCG66AGBSA092N DDR5 5600MHz CL46 1.1V SO-DIMM's laying around, so I thought I'd try it. Some comparison's in this thread, and this one.
The cheapest DDR5 SO-DIMM to U-DIMM adapters I found were around $13.50 per, most of them without reviews. The ones with good reviews were closer to $25. I opted for the cheaper ones.
I put them in my daughter's Ryzen 5 9600X computer, and they showed up at 3200MHz in the BIOS.
I tried 5600MHz, and it wouldn't boot.
5400MHz? Nope.
But then, 5200MHz, and it booted into BIOS!
I ran a quick CPU-Z benchmark, and it scored well.
CPU-Z Single: 837.4
CPU-Multi: 6380.3
With everything set to AUTO in BIOS, I'm getting 42-42-42-84-125 at 5200MHz with SK Hynix HMCG66AGBSA092N DDR5 5600MHz CL46 1.1V memory.
While it doesn't look very cool, they perform well. I've got enough of a light show where my daughter wouldn't even notice, anyway.
But, it's not over, yet!
Y-Cruncher was unstable.
I had to set it to 5000MHz in order to pass Y-Cruncher. Benchmark Pi, Multi-Threaded (no SMT). Total Computation Time = 75.932 seconds.
Re-running CPU-Z gave me a better score.
CPU-Z Single: 839.9
CPU-Multi: 6420.5
Here are the comparisons to a desktop DDR5 6000MHz 2x8GB CL38 Kit at EXPO1:
Here are the comparisons to the same desktop DDR5 6000MHz 2x8GB CL38 Kit at EXPO1 and 6400MHz:
Total comparison scores:
DDR5 5600MHz SO-DIMM at 5000MHz
CPU-Z Single: 839.9
CPU-Multi: 6420.5
Y-Cruncher: 75.932 seconds.
DDR5 6000MHz U-DIMM at EXPO1
CPU-Z Single: 842.9 (0.36% faster)
CPU-Multi: 6412.9 (0.12% slower)
Y-Cruncher: 74.615 seconds (1.75% faster)
DDR5 6000MHz U-DIMM at EXPO1 and 6400MHz
CPU-Z Single: 844.0 (0.49% faster)
CPU-Multi: 6411.4 (0.14% slower)
Y-Cruncher: 73.909 seconds (2.70% faster)
The bandwidth might be 600MHz slower, but the timings scale to where they still perform well. As far as Y-Cruncher and CPU-Z is concerned, you don't really lose a lot of performance.
Good enough until the RAMpocalypse is over.
Update (06/08/2026), (07/05/2026), and (08/14/2026):
It seems that certain motherboards are better than others with the adapters. AM5 motherboards fare better than LGA 1700 DDR5 motherboards, and the quality of the CPU can also make a difference.
1. Worst motherboard - Gigabyte B760M C Rev 1.3, I got it cheap from a pre-built, so I knew it wasn't going to do well. It took a long time to get it stable, because if the timings were too slow, it wouldn't boot. So I had to keep dropping the frequency while lowering the timings until I got it stable. Ended up stable at 4400MHz at 56-56-56-112-168-2T.
2. 2nd worst motherboard - ASUS TUF GAMING Z790-Plus WiFi. Stable at 4200MHz at 26-26-26-52-76. Worst speed so far, but fastest timings.
3. 2nd best motherboard - ASUS B850 MAX GAMING WIFI W. Stable at 4600MHz at 30-30-30-60-90.
4. Best so far - Asus PRIME X870-P WIFI. Used in the initial tests. Final stable settings were at 5000MHz at 46-46-46-92-138.
All the AM5 motherboards I've tried with a 9850X3D can get stable at 4800MHz at 42-42-42-84-128. However, I found it faster at 4400MHz at 32-32-32-64-96. One was even faster at 30-30-30-60-90.
Also, if using a single stick, you can get faster frequencies and timings than with two sticks. Using 4 sticks hasn't really worked, and I think to make it work, I'd have to keep dropping the speed and increasing the timings.
Conclusion:
So far, I've not had a problem with these in the long run. I've also been able to get them working on every DDR5 motherboard I've tried, although I haven't yet tried on an LGA1851 motherboard. It appears that in general, so far, AM5 motherboards have been more successful in regards to speeds and timings than LGA1700 motherboards. Yes, there's some performance loss, but it's really not that much, and I can't even tell except from benchmark results. There's no crashes, stuttering, slowdowns, or anything of that sort.
The cheapest DDR5 SO-DIMM to U-DIMM adapters I found were around $13.50 per, most of them without reviews. The ones with good reviews were closer to $25. I opted for the cheaper ones.
I put them in my daughter's Ryzen 5 9600X computer, and they showed up at 3200MHz in the BIOS.
I tried 5600MHz, and it wouldn't boot.
5400MHz? Nope.
But then, 5200MHz, and it booted into BIOS!
I ran a quick CPU-Z benchmark, and it scored well.
CPU-Z Single: 837.4
CPU-Multi: 6380.3
With everything set to AUTO in BIOS, I'm getting 42-42-42-84-125 at 5200MHz with SK Hynix HMCG66AGBSA092N DDR5 5600MHz CL46 1.1V memory.
While it doesn't look very cool, they perform well. I've got enough of a light show where my daughter wouldn't even notice, anyway.
But, it's not over, yet!
Y-Cruncher was unstable.
I had to set it to 5000MHz in order to pass Y-Cruncher. Benchmark Pi, Multi-Threaded (no SMT). Total Computation Time = 75.932 seconds.
Re-running CPU-Z gave me a better score.
CPU-Z Single: 839.9
CPU-Multi: 6420.5
Here are the comparisons to a desktop DDR5 6000MHz 2x8GB CL38 Kit at EXPO1:
Here are the comparisons to the same desktop DDR5 6000MHz 2x8GB CL38 Kit at EXPO1 and 6400MHz:
Total comparison scores:
DDR5 5600MHz SO-DIMM at 5000MHz
CPU-Z Single: 839.9
CPU-Multi: 6420.5
Y-Cruncher: 75.932 seconds.
DDR5 6000MHz U-DIMM at EXPO1
CPU-Z Single: 842.9 (0.36% faster)
CPU-Multi: 6412.9 (0.12% slower)
Y-Cruncher: 74.615 seconds (1.75% faster)
DDR5 6000MHz U-DIMM at EXPO1 and 6400MHz
CPU-Z Single: 844.0 (0.49% faster)
CPU-Multi: 6411.4 (0.14% slower)
Y-Cruncher: 73.909 seconds (2.70% faster)
The bandwidth might be 600MHz slower, but the timings scale to where they still perform well. As far as Y-Cruncher and CPU-Z is concerned, you don't really lose a lot of performance.
Good enough until the RAMpocalypse is over.
Update (06/08/2026), (07/05/2026), and (08/14/2026):
It seems that certain motherboards are better than others with the adapters. AM5 motherboards fare better than LGA 1700 DDR5 motherboards, and the quality of the CPU can also make a difference.
1. Worst motherboard - Gigabyte B760M C Rev 1.3, I got it cheap from a pre-built, so I knew it wasn't going to do well. It took a long time to get it stable, because if the timings were too slow, it wouldn't boot. So I had to keep dropping the frequency while lowering the timings until I got it stable. Ended up stable at 4400MHz at 56-56-56-112-168-2T.
2. 2nd worst motherboard - ASUS TUF GAMING Z790-Plus WiFi. Stable at 4200MHz at 26-26-26-52-76. Worst speed so far, but fastest timings.
3. 2nd best motherboard - ASUS B850 MAX GAMING WIFI W. Stable at 4600MHz at 30-30-30-60-90.
4. Best so far - Asus PRIME X870-P WIFI. Used in the initial tests. Final stable settings were at 5000MHz at 46-46-46-92-138.
All the AM5 motherboards I've tried with a 9850X3D can get stable at 4800MHz at 42-42-42-84-128. However, I found it faster at 4400MHz at 32-32-32-64-96. One was even faster at 30-30-30-60-90.
Also, if using a single stick, you can get faster frequencies and timings than with two sticks. Using 4 sticks hasn't really worked, and I think to make it work, I'd have to keep dropping the speed and increasing the timings.
Conclusion:
So far, I've not had a problem with these in the long run. I've also been able to get them working on every DDR5 motherboard I've tried, although I haven't yet tried on an LGA1851 motherboard. It appears that in general, so far, AM5 motherboards have been more successful in regards to speeds and timings than LGA1700 motherboards. Yes, there's some performance loss, but it's really not that much, and I can't even tell except from benchmark results. There's no crashes, stuttering, slowdowns, or anything of that sort.
Last edited:
