LatencyMon Benchmarks

Post your awesome projects here for everyone to marvel at!
Joe-W4WT
Posts: 52
Joined: Sun Apr 09, 2017 5:27 pm
Location: Cumming, GA

Re: LatencyMon Benchmarks

Postby Joe-W4WT » Tue Aug 07, 2018 2:06 am

Hi Gary,

As you can see in my picture, cores 5, 6, & 7 are obviously not following cores 0 thru 4 and are running at a lower cpu rate so the Affinity change worked for me.

If you only moved the programs you wanted to core 6 and did not prohibit other programs from using Core 6 then that may be why you didn't see a change. You can't just move PowerSDR (and your other stuff) to Core 6 without removing other programs from using core 6 too, otherwise you haven't accomplished anything and in fact you may have hurt the apps you moved to core 6 as they now have only one core to operate on along with all the other programs that are using core 6 and all other cores.

73,

Joe W4WT
K9RX
Posts: 209
Joined: Fri Apr 14, 2017 3:47 pm

Re: LatencyMon Benchmarks

Postby K9RX » Tue Aug 07, 2018 12:46 pm

Joe,

Agreed ... however there are LOTS of things running ... I don't know how programs that do not run multi-thread (if that is the right term for actually using multi-cores by design) end up using a core ... if they are only typically on a core - then unless there is some randomness on how they're assigned I'd think that there might be a tendency to overload the lower core numbers - that was my assumption. So I selected the last core (which by the way on this i5 is actually core 3, not 6). That's what I was going with. But first and foremost i need to figure out a way to see my cores as TM is only showing the CPU results - I know I USE to be able to see cores - not sure why they removed/moved it on Windows10.

Also apologies for not being explicit... when I said it didn't change I meant there was no obvious bad result - it ran as normal with no visible (or audible) effect.

I have since played with it a good bit more - this time looking at turning on the resampler which previously never resolved and was always hunting. With both PSDR and WSJT on the same core - nothing else changed, I was able to see what appears to be an improvement in the resampler - there were times with certain settings, changing the Buffer size and the Buffer latency value/manual(auto) where it would settle down - only a single digit changing for a good couple of minutes ... but ultimately it too would 'run-away' with a sudden rapid change in the ratio and huge jumps in the OF/UF numbers ... so I don't know if that is 'better' or if it is just different... it was indeed different... I realized later that I really need to put VAC-M on the same core as well - but I've no clue where to find it in the services or details tab in order to change the affinity. There's nothing obvious - I could use help here if someone knows what its 'details' would be.

This is just an experiment - I agree with you that moving things off of the "radio system" core would be ideal - but unless there is a way to automate that we're just adding to the (already) many tasks required at computer/radio startup time ... I'm currently looking at a slightly less onerous set of changes to see if that improves things noticeably.
K9RX
Posts: 209
Joined: Fri Apr 14, 2017 3:47 pm

Re: LatencyMon Benchmarks

Postby K9RX » Sun Aug 19, 2018 2:31 pm

so that new computer that i have ... it is an i7-8700 3.2Ghz ... it ran a 14800 on CPU/Passmark test 2 weeks ago or so ... today when I run it it is anywhere from 11500 (tops) down to 10000! WHAT CHANGED? I can't figure out why it would drop so much. That is a huge difference and its now running about as fast as my 5 year old i7-4770. The "nominal" for this new CPU is 15200. The only thing that I can think of that changed is that I installed a Samsung NVMe SSD (it is blazingly fast). I probably installed some programs since then - I'm sure i did - but I've disabled any that come up in the STARTUP and don't see any change in the results - still 11K'ish. I have searched on the internet to see if there's anyone that has had Passmark results that changed this drastically - nothing found. Has anyone seen this (and hopefully resolved it)?

and yes, It is using all 12 processors.

Gary
K9RX
User avatar
w-u-2-o
Posts: 1172
Joined: Fri Mar 10, 2017 1:47 pm

Re: LatencyMon Benchmarks

Postby w-u-2-o » Sun Aug 19, 2018 3:26 pm

I have seen similar changes in the Passmark ratings of my PC. It has never been as fast as the first week I had it running. It would seem that as Windows gets clogged up with the detritus of normal, daily operations, the system does tend to slow a bit. However my changes are not nearly so dramatic as yours.

73,

Scott
K9RX
Posts: 209
Joined: Fri Apr 14, 2017 3:47 pm

Re: LatencyMon Benchmarks

Postby K9RX » Sun Aug 19, 2018 4:01 pm

I'd "accept" that except for when I tested this computer, the 5+ year old i7-4770, it's rating is spot on for this processor results as shown by PassMark (showing hundreds of results). I didn't do anything to this computer before I ran the test other than start it up. Also we're talking 19 days since I ran that first test, not months...

frustrating.

I'm tempted to return it, the result I'm getting is in the bottom 0.1% of all submissions!

Gary
User avatar
WA0VY
Posts: 70
Joined: Sun Apr 09, 2017 3:41 pm

Re: LatencyMon Benchmarks

Postby WA0VY » Sun Aug 19, 2018 4:52 pm

FWIW, my Xeon is still performing about as it did originally and still at or above the average results on Passmark with very minimal OC-ing. I have not seen any deterioration in performance over the last year worth noting. I even ran Passmark with PowerSDR, ASI, Omnia and other software running and found only a slightly reduced result. Gary, I can't imagine what would be knocking down your performance to that degree.
73 Brent WA0VY
K9RX
Posts: 209
Joined: Fri Apr 14, 2017 3:47 pm

Re: LatencyMon Benchmarks

Postby K9RX » Mon Aug 20, 2018 12:29 pm

so I spent the whole d#$* day working on it ... I went back to the original HDD alone, tried the SSD alone... tried various settings etc ... always the same results - around 11500 ish when the first test gave me a 14480 and the nominal is around 15250. I even saw one result that was at 10,700! That is a 30% drop. I spent over an hour on the phone with HP support. Of course they want to dismiss it. But the differences in performance are not mind blowing when you get up there as we all know - but you DO pay for those differences!

I ran PT on this computer - a 5+ year old i7-4770 ... results were right on the mark for this CPU and not much worse than what I'm getting on this new one... then I ran it on my current radio computer, an i5-4460 - again right on the mark, actually about 8% higher than nominal. This was with no changes, nothing turned off from the start menu - not even removing any of the other "stuff" running like the virus program etc. So obviously the test is as they say it is - designed to test the CPU performance with little impact from memory/disk/video and other things running (within reason I'm sure). These are both Dells, the new one is an HP.

I finally did an OS recovery using the original HDD. STILL only 11000. SOMETHING has changed. Something is not working right. I'm going to see if I can exchange it first - if that doesn't work then I'm going to get a refund. It might well be that the old axiom "you get what you paid for" is rearing up ...

Gary
[the admin at Passmark has suggested a good number of things to check - sadly none applied to my situation - he was there though and helpful!]
User avatar
WA0VY
Posts: 70
Joined: Sun Apr 09, 2017 3:41 pm

Re: LatencyMon Benchmarks

Postby WA0VY » Mon Aug 20, 2018 9:23 pm

As a post script. I ran a full Passmark benchmark and found significant deterioration in my GPU 3D score and hard drive score which are now far below normal. Again, no noticeable deterioration in GPU 2D, CPU or memory scores. I am contemplating doing a full burn-it-down and reload from a bare hard drive. I have periodically done that for years and it has always helped eliminate the glitch build up, which the signs are now pointing to.

Gary, you said you did an "OS recovery." Not sure what you mean by that. Did you take it all the way down to a reformatted hard drive level?
73 Brent WA0VY
K9RX
Posts: 209
Joined: Fri Apr 14, 2017 3:47 pm

Re: LatencyMon Benchmarks

Postby K9RX » Tue Aug 21, 2018 1:44 pm

Hey Brent - I fancy myself a history buff - but I can't tell who the picture avatar is ... ?

So ... I just did an OS restore. I believe all computers now ship (if pre-built) with a partition on the HD with Windows files/installer. So I just ran their utility to do that. It bought it back to where it was when first received.

That stated: I think I know what it was. So I had it new - I did the test and got near nominal results for the CPU. 14840 vs. nominal 15250 or so. Then I started to go through Windows to find things set for a balance between power and performance (which was everything setable) and set all that I could find to performance... I installed the programs needed ... finally installed the new NVMe SSD. Oh - and to be sure I went to HP's site for this desktop and found any and all drivers that were not the latest and installed them. This included the BIOS.

It was at that time - checking things out - setting up the programs, logger, audio, Dtime, etc etc that I re-ran the performance test and got the low CPU results. The results were best case down 22% and worst case down 30%.

I found out yesterday, on the HP Forum, someone posted that apparently Intel introduced a bug in their processors in their never ending quest for faster performance - that made them hackable. To fix it you needed to ALSO repair (upgrade) the BIOS. I was not aware of this when I upgraded the BIOS above. Apparently this "fix" caused performance degradation to the tune of - wait for it - as much as 30%! Some were reporting 50% reduction in certain games etc.

So since i went back to the original shipped computer setup with only the bios having been changed and it worked pre and didn't post the bios change - it had to have been the BIOS upgrade that did it.

By the way I bought this from Staples, had run past the allowed 14 days for returns (at about 24 days) due to issues getting answers about the SSD drive and then this issue with a drop in performance where HP hadn't a clue with hours spent with their support. I was denied the ability to return it - so I called their customer service and talked to a supervisor. I explained the situation and she was kind enough, it was reluctantly, but kind enough to allow me to a) do an exchange for a new unit to see if there was something wrong with this one and b) if that doesn't work get a refund. So kudos to Staples. I normally would not have pushed it so far out except I thought it was a 30 day exchange policy.

Gary
K9RX
User avatar
WA0VY
Posts: 70
Joined: Sun Apr 09, 2017 3:41 pm

Re: LatencyMon Benchmarks

Postby WA0VY » Tue Aug 21, 2018 2:42 pm

Yes, yes, I remember the hack issue and the fix that would slow down the CPU. Good catch. At the time I heard the news, I checked to see if my Xeon was problematic and found it was not. So that may explain why I saw no slow-down on my CPU but, like Scott, I did see deteriorating scores over time on other systems.
73 Brent WA0VY
User avatar
w-u-2-o
Posts: 1172
Joined: Fri Mar 10, 2017 1:47 pm

Re: LatencyMon Benchmarks

Postby w-u-2-o » Tue Aug 21, 2018 11:23 pm

Ah yes, the processor vulnerability bug. I have a custom build using an EVGA motherboard. I monitor their forum, and when the released the new BIOS with the upgraded microcode to address the bug people reported a substantial drop in SSD performance. I did not upgrade and accepted the security risk instead.

I suspect that anyone running the new microcode is screwed, whether due to an upgrade or a brand new purchase.
Bryan W4WMT
Posts: 18
Joined: Sun Apr 09, 2017 10:12 pm

Re: LatencyMon Benchmarks

Postby Bryan W4WMT » Wed Aug 22, 2018 9:48 am

Is that the “speculative branching” microcode vulnerability that everyone was freaking out about not long ago?
User avatar
w-u-2-o
Posts: 1172
Joined: Fri Mar 10, 2017 1:47 pm

Re: LatencyMon Benchmarks

Postby w-u-2-o » Wed Aug 22, 2018 10:09 am

Afaik, yes.
K9RX
Posts: 209
Joined: Fri Apr 14, 2017 3:47 pm

Re: LatencyMon Benchmarks

Postby K9RX » Thu Aug 23, 2018 3:27 pm

Gentleman: Hopefully this will be the last report to this thread.

So - I got the replacement i7-8700 in, this is a Pavillion 590-p0070. The first one had reasonable PassMark PerformanceTest marks for the CPU coming in at 14840 (the "nominal" over what appears to be hundreds of submissions for this processor is about 15200). But then after doing some stuff to the computer it dropped by as much as 30% and never again showed any closer than -21%. The suspicion was, as submitted by someone here (thank you!) that it was caused by my upgrading the BIOS to the latest version. There is a bug in Intel Processors that make them 'hackable' referred to as Spectre and/or Meltdown (2 different attach methods I believe). The "fix" requires the BIOS to be upgraded as part of the fix - and the result of this has been reported as performance hits to the tune of, as much as, 30%!

This is where I say BINGO!

So I exchanged that unit for a new one. That unit on first test just after the very first power up showed not much better at 12600 ... a huge disappointment as its closer but still NOT the performance you pay for.

HOWEVER after both cycling power several times and running it i noticed that the HDD was running constantly at 100% ... so I can only assume that Windows was doing HUGE updates while the computer was running (the NETWORK was also going up and down as if things were being downloaded). So I let this run waiting for the HDD to go to 0. This took a good HOUR. Finally I checked it again after all this and BINGO - 14941!

so the caveat here is - on new computers do NOT run this test right away. Let the computer come up, run for several minutes, restart it and repeat several times, then let it sit connected to the web for at least an hour or two... then try the test.

I'm going to now start loading in the needed programs and changing the Windows performance settings - I'll check the CPU results often along the way just in case it gets hit again so I know the before and after and cause.

The final Passmark results, this is with the onboard video (which apparently needs a driver update as it failed part of the test due to this) gave me 14941 for the CPU and 17700 for the disk! The disk being now the Samsung 970 EVO NVMe PCIe 4-ch.... it ROCKS big time and is at the 99% of all systems. Memory came in at 3043 which was I believe 95%.

Gary
K9RX

Return to “PC's & Homebrew Hardware”