Recent content by twin_savage

The Samsung 840 evo situation from years ago was interesting because the drive's controller was just being overly sensitive in applying all the extra degraded read tricks when they weren't truly necessary due to an errant calibration value. Presumably those drives would slow down to the same...

Yes, that's the one. It goes off the rails a couple of times but in general there is some good information on there.

The behavior on the consumer drive you are experiencing is almost certainly NAND cell charge decay. The SLC cache recovery on these consumer drives happens within a matter of minutes to maybe an hour for the very largest SSDs with a high SLC cache ratio. I'd expect the consumer drive to...

This may be true for Broadcom, but hardware raid cards and even HBAs from more helpful vendors let you attach to NVMe disks with x1, x2 or x4 PCIe lanes per SSD without a backplane; just don't expect them to be passed through as completely native NVMe (some of the NVMe featureset is going to be...

I can confirm that they will not. That being said, SlimSAS 8i S‌FF-8654 to 8 X SFF-8639 U.2 cables exist but I doubt the Broadcom HBA would be compatible with them since Broadcom doesn't offer the cable for sale themself.

I just realized there are no SNC settings in BIOS, SNC is always 1 for this motherboard and CPU combo, so of course virtual NUMA and UMA-based clustering work. For those interested in how Virtual NUMA and UMA-Based Clustering affect memory latency via MLC: Virtual Numa Disabled, UMA...

W790 BIOS settings on this are confusing, it makes it sounds like UMA modes cannot be enabled based on "the current system configuration" but when I enable them and try different clustering schemes I get reproducible differences in memory performance: It almost sounds like UMA-based...

I just ran passmark and updated my post, the numbers seem kind of strange, it reports 57ns latency which is very optimistic. I included the MLC benchmarks that show true single thread memory performance aswell.

This thread has a very interesting topic for me because my primary compute need is solving very large system of equations that scale very strongly with memory bandwidth and latency and take hundreds to thousands of hours on current SPR systems to solve. WRT the original post that spawned into...

The Intel MLC tool is compiled for Linux and can check latency; and it's read test on Linux also produces numbers identical to AIDA64's read performance on Windows.

Here's my AIDA64, Passmark and MLC results:

Here's the output: IOMMU group 17 05:0a.0 PCI bridge [0604]: Advanced Micro Devices, Inc. [AMD] Device [1022:43f5] (rev 01) 09:00.0 PCI bridge [0604]: ASPEED Technology, Inc. AST1150 PCI-to-PCI Bridge [1a03:1150] (rev 06) 0a:00.0 VGA compatible controller [0300]: ASPEED...

For reference, the 41 watt power figure on the H13SAE system was with a 850watt corsair hx850i platinum power supply. I was previously running the exact same system on a 1500w Silverstone ST1500 silver power supply (because I thought I needed more 5v rail amps) and the idle draw was 61watts...

PCIe traffic doesn't contain routing information so you can't really have multiple hosts sharing a pool of devices outside of some very very specific architectures that are tailor made for businesses. The Dell VRTX is probably one of the more flexible examples of this, but it is only PCIe 2.0...

The non-chiplet LGA-4677 Xeons are alot better than the chiplet ones with idle draw, I remember hearing third hand that they are ~40 watts idle, that is CPU power draw, not whole system. E-cores are actually fairly potent, at least compared to CPUs of the recent past; If you wanted to go all...