Patriot Viper Steel 4000-19 Samsung B-Die performance data Overclocking outcomes recorded
Patriot Viper Steel 4000-19 Samsung B-Die performance data Overclocking outcomes recorded
Test System; MSI MPG B550 Gaming Plus AMD Ryzen 7 5800X @PBO +200 MHz CO -20 all cores Patriot Viper Steel 4000-19 2x8 Samsung B-Die TestMem5 1usmus_v3 config 10 Cycles no error. Prime95 Large FFTs no WHEA. Prime95 Small FFTs no WHEA. MemTes86 no errors. AIDA 64 51.2ns. These are the most precise numbers for my build. I have poor IMC, can't exceed 2000 FCLK; actually I can, but it requires a 1.25 SoC and I wouldn't choose it for regular use—it's not needed at 67 MHz and 33 FCLK. I achieved 51.2ns in AIDA64. That's sufficient for me. Wishing success to all who overclock. By the way, my CPU is overclocked to PBO2 +200 MHz and -20 CO for all cores.
Impressive outcomes from this setup. My current rig runs at 4000 16-16-16-32 1.45v on a 10850k board, but I would have preferred a gigabyte board for the t-topology. I plan to adjust all my OC settings once I upgrade to a proper case with a 360mm aio, which could make the Z490 gaming edge even more surprising.
Providing the correct voltage works best for this chip. If temperatures stay below 60°C, it shouldn't be an issue. I shared a voltage setting of 1.57V using the stock LLC. You can also adjust the timings, as some—like tRCDRD, tRFC, tRC, and tCL—are sensitive to voltage changes. I can lower the tRFC value further, but it might not be ideal. A voltage of 295 is acceptable for me. Poor IMC can cause high ns Unluckiness, but overall I'm satisfied with 1967. By the way, your results are great!
No problem man! Thanks! The FCLK is not stable at that speed, errors start when I pass 2000 1:1. The memory was getting about 1.6v I think.
I'm really struck by the 1967MHz FCLK. None of the Ryzen 5000 models I've tested reach more than 1900MHz without needing scaling adjustments. My 5900X, for instance, only runs about 26 seconds in Y Cruncher 1B at 3800MHz CL16-13-13-21 dual rank, but it climbs to over 5 minutes with a bit more relaxed timing. You're pretty lucky the CPU performs this way if it doesn't.
It seems this B-Die at 4000mhz on the CL14-14-14 platform should be achievable, especially on Intel's setup. The main issue with AMD is the Fclk cap; even their 5000 series aren't stable at that frequency, and my 2700X works at 3933mhz but not at 4000. There’s no solution for decoupling it either. A full B-Die can reach up to 4000mhz on CL12 or CL11, but my A2 single won’t. The 4200 CL16/17 actually gave better 3D scores.
They’ll go for a full 3600 at 12 bars, while my black and white designs refuse to include the C12 version.
I used tRCDRD to 14, but it posts inconsistently. It seems I need higher voltage for 14, possibly. The current voltage is 1.62V and it remains unstable. I lowered the voltage to 1.57V and it worked, but at 1.56V it becomes unstable again.