Recent content by MichalPL

short answer: 1600MHz If remember correctly (now using Lenovo 8x 8894v4 - and thinking about TR7000 single CPU, overclocked ;) ): G8 (DDR3) with v2 CPU's: was always max of 1600MHz (and 1866MHz in the mode that you have not 8 channels(4 SMI channels) but 4). G9 (DDR4) with v4 CPU's: was always...

New W9-3495X quite high in passmark top charts

my mistake

Talking about "true" ECC DIMMs ? EDIT: DDR5 unbuffered and DDR5 registered is different. DDR3 and DDR3 ECC can work on same motherboards (LGA1366/2011/some 2011-v3) and CPU (for example E5-2678 V3). DDR4 and DDR4 ECC can work on same motherboards (LGA 2011-v3/3647/...) and CPU (for example...

ok almost 3900 points sound really good (@4.8GHz) :) closer to the Intel 12gen, than Ryzen 5000 - ok this is probably it - i9 with proper number of pcie lanes

ok - 100% right, with the second result it is much better. https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+w9-3475X&id=5315 now it's better: 65891 -> 70850 3678 -> 3725 if they (passmark) do just average of the all results, the second result is quite high: 2*70850-65891 = 75809 passmarks...

Integer score or floating point score is a multithread only. single core probably have same/similar tests running just on 1 core. 2012 to 2020 November (Release of Ryzen 5000), it was like this. all Xeons starting form Westmere up to ~10gen i9 for large socket (2011/2011v3/2066 and 36xx-...

No - it will be 1:1 to Zen3, and slower than Zen4 (who exist only in non-Treadripper parts) 24 core AMD (5965WX, Zen3): Extended Instructions 66,924 Million Matrices/Sec 32 core AMD (5975WX, Zen3): Extended Instructions 81,382 Million Matrices/Sec 36core Intel Xeon (w9-3475X...

I hope more results shortly, and yes hope it was tested on 1dimm of DDR5 ;)

Yes, this is a bad thing :/ slow in multi-core Good thing is it finally the fastest single core performance CPU (still like Ryzen 5900X) with massive number of PCIe lanes. so in short... maybe fast enough to build ZFS NAS based on NVMe PCIe 4.0 who can saturate 2x 100GbE when copy 1 large...

first results of the Intel Xeon w9-3475X - slightly slower than predicted Intel Xeon w9-3475X - 65891 points! https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+w9-3475X&id=5315 Single Thread Rating: 3678 AMD Ryzen Threadripper PRO 5995WX - 96892...

Yes, but I don't recommend PCIe x8 for 2x100G, because exactly like Freebsd1976 wrote it will be about 80GbE (display as 100GbE link of course), what is not super bad for 40GbE replacement but it's not true 100GbE. for 2x100GbE PCIe 3.0 x16 is also too slow (will do "160GbE") first good...

on ConnectX-3 working @ 40GbE i was able to saturate the link on samba under windows, 4.42GB/s. I didnt try this @ 56GbE but should be faster. I see right now pcie x8 for $100 and x16 for $159. (50GbE variant need to flash to 100GbE)

I was using for a short while Cosco 3164Q (backthen cost about 1000 EUR) but switched to much faster (and cheaper) Celestica 100GbE -but less ports, and Mellanox 56GbE. i am using: ConnectX-3 for 40G ConnectX-4 for 100G only ConnectX-5 for 2x100G only and having experimental 200G but no...

I think even cheaper bought 6 switches for about 380-460 usd, all working good (with fixed atom cpu) Plus cards (required flash to work as 100GbE) for 100-120 EUR but long time ago, plus big transcievers box for about $8each. Now maybe 400GbE will be possible somehow, and pcie 4.0/5.0 ;) but…...