I.nfraR.ed

I can confirm! Very shady practice. I'm really disappointed. Does not affect me, since I use blockers, but... After disabling uBlock Origin on hwbot forum, CPU goes to 100% on all cores.

The most enjoyable benchmark to explore scores on hwbot...not. No screenshot, no rig pictures, no readings on CPU clocks, mem speed and timings, just a number in the rankings. Fore sure very entertaining for the public. Said noone ever. Nothing against you, Roberto. Don't get me wrong.

I would suggest to use Asus/Asrock tools to patch your own version of Windows. I'm using a stripped down SP1 and have integrated the drivers with NTLite, but the tools should automate the process. Can't share it though, for obvious reasons.

Exploring obscure combinations of hardware sounds interesting, but that would probably rule me out, too. Whenever I bench for some competitions, I'd like to combine this with hw points for the rankings and this almost always means the best possible mb-cpu-gpu combo. So buying some motherboard with lesser performance is not the ideal scenario for me and would probably skip it.

@Leeghoofd Here's the threat from my side. If Alby doesn't participate in the next comp, I won't participate either* Always good to kick his old ass with some AMD scores *Disclaimer: Alby's participation doesn't automatically mean I will too.

Unfortunately this result is bugged. Not sure how, but subtests are inflated more than 3-4 times, even though it appears as valid on FM site.

Good find! Seems I was smart enough (read "haven't benched anything in the last several months") to not bother about these benchmarks. Perhaps I'll now move up in the rankings without benching anything, yay!

Just speculating, but I'm with others saying it is a VRM problem. I have something similar on an AM2+ board which only boots lower power single-core CPUs and the reason for this is a blown mosfet.

Yeah, NCLK/Infinity Fabric, although I can actually control the NCLK with this bios, but it really depends. On some bioses it was just auto and this divider was bugged, so 1:8 was the way to go ( I assume this is the reason you ran 1:8), however in my case I don't see a performance penalty if everything is set correctly and this higher divider allows me to use much lower refclk, which seems to be the limiting factor on 1:8 with my setup. I was expecting to see some difference between even and odd dividers, but they are really close in terms of performance. Might play a little bit and mod the bios to unlock some hidden features (if any), which are available on older beta bioses, although they were quite buggy. vNB is limited to 1.33V, while I've seen much higher on older ones, although I try to be gentle with volts, because higher is not always better, especially on APUs. Not going to run 3D, but I think it is clear to me what is needed to get a high score. Haven't seen bigger boards though, this one has rather limited bios options. Guess it might be platform specific, have to dig up some old reviews to see.

I"m dropping channel roughly at that mem frequency, too. Independant from refclk. This 1280 was hit or miss, sometimes it boots dual, sometimes single, so I have to turn off and then on until it sees the second channel. Freezing the CPU might help. Score could be better for sure and it is not that hard of a platform to figure out, but it was never popular, so not many people benched it. But even without optimizations the current #1 is easy to beat when you dial in everything. Maybe I should spend some time with LN2, too. This chip is the highest validated on the bot.

New Llano top score: I.nfraR.ed`s SuperPi - 32M score: 15min 49sec 812ms with a A8-3870K Current one is easy to beat, indeed.

Clearly your run is without wazza. There goes my result and it could be even faster, although this IMC/MB/CPU combo is on the edge of stability as far as mem goes. Without wazza I was more than 1sec faster than your system. Some things worth mentioning in case you wonder: - This higher divider didn't seem slower than 1:8 and it is needed in my case, because the board starts loosing the disks after 154MHz and doesn't even post at 160. Might be the board, might be the CPU. Subtimings that can't be controlled manually from bios seem to be exactly the same as on 1:8 div. - CAS7 is possible, but it is not faster than CAS8 at least on this board. tRCD 11 is highly unstable, I think that might have better luck with BBSE, but this board is very picky about memory and I used PSC. - Overall the current number 1 is much slower and very easy to beat. Not many people bothered to run this platform, though, so this is one of the reasons. PS: Second image is without wazza, first one is my current best. Well, this is offtopic, but I don't see many people running 32M on Zen anyway.

Finally got it (the FM1 board) to post after several tries to force backup bios o reflash main bios. Managed to do it by shorting GND to #HOLD on the M_BIOS, which basically disables main bios initialization and forces backup bios load, but it didn't work first try. I've had the same on other Gigabyte boards. Now off to installing OS and some benching

About AM4: It's either the cpu or the board, because I've tested the memory and it works on a Intel system. And I sold my other AM4 board, so can't test the cpu. As for the RAM - I have 3 remaining B-die kits - one FlareX I bought from the store and 2 TridentZ binned by websmile, but my other old TridentZ 3600C17 could do it as well. So it is the IMC, I believe. Seems to be ok on cold too (only tested on chiller though). Sadly the Llano system doesn't post I haven't touched it the last 5 years, will try to do some contact cleaning. Really wanted to spend some time on it.

Since my Zen platform is no functioning now (I think a metal washer from the cooling block shorted something), I'm gonna play some Llano. I only hope the 1800X is alive, since it can do DDR4-3800+ tight and has pretty good cores as well, but haven't tested it on LN2 yet.

I would be interested in socket 7, socket A, socket 939, AM2(+) AMD CPUs, but it would take me a lot of time to compare your tables with mine and tell you what I'm missing. Would definitely buy many CPUs depending on price to fill missing models in my collections. Pretty interested in all A64 FX CPUs, although I don't have FX-60 only. Maybe some of the higher rated Opterons, others I have. Interested in NF4 SLI-DR Expert Venus as well - the only board I don't have from DFI's NF4 lineup.

Hmm, maybe things have changed, but I remember it was working this way, because it happened to me as well in the past. But I haven't validated anything for quite some time. Maybe my memories have faded away, don't know.

You should wait, but I don't know what is the period set on the site. Just try again after a while, maybe tomorrow - it should validate.

Score is 1165, not 1439.

Here's mine. Got a lot faster runs yesterday (maybe in line with what chew* posted, although at much higher mem speed), but didn't finish due to some low voltages I was experimenting with and I had no software installed to change them runtime in OS. Was easily 2-3 seconds faster than this (1st loop - low 23s, 2nd - mid 43s, 3rd 1m 03s, etc). Will keep trying when I have time, but still can't figure out all the odd things.

Got it down to 8.17.xxx with mem at 3600 12-12-12 without wazza, BUT... Simple reboot, same settings, nothing changed and boom - 10 seconds slower. And that is much more frequent than sub 8.20. Maybe something internal is not triggered every time (or almost never) or it is some other sort of bug, still can't figure it out. Will upload the screenshot in the evening.

Ah, ok, pci-e sata card makes sense, didn't think about that.

Ok, but have I missed something or there are Zen drivers for XP? Last time I checked it was not supported and I've tried myself, but could not install it, nor run an already installed OS from another platform. Otherwise I agree about the XP for 32M, the source iso matters as well.

XP with Ryzen?

As far as I know the only difference is the heatpipe and the different heatsinks on the VRMs and NB. PS: Judging from pictures, M2RS does not have QCODE display, also has different inductors, maybe something else. Edit: Apparently, the southbridge is SB600 vs SB750, didn't know that.