(Anti) SFF fun house

[Mussels]

temp related could make sense, if the RAM literally freezes/pauses for a moment, then wakes up literally a second later - one brief bit of stutter and its cool

whys it erratic? cause that one ram module simply may not be the one in use, the next time

 

[Valantar]

Hm, that is really interesting. I've been seeing some ... possibly similar weirdness on my system, in particular packet loss and latency variation warnings in Rocket League (it's the only online game I play with some regularity). Seems to happen regardless of the internet connection quality. This is with my 3200c14 B-die though (running at DOCP, still haven't bothered to OC it any). I might stick my old TridentZ kit in here just to see if anything changes.

As part of troubleshooting this, I've discovered that my system seems to get some pretty bad DPC latency issues after one or more sleep cycles, particularly when left asleep overnight. After a reboot everything is fine though. Any chance you could run LatencyMon with your previous kit after a sleep cycle or two to see if there is any relation? (I've been in contact with ASRock about this, and they say they can't reproduce it.) My RAM never exceeds ~45°C though.

 

[Mussels]

There was that thingy with ryzen and the WHEA errors causing the USB ports to reset, because of errors on the PCI-E bus (usually caused by PCI-E 3.0 risers mostly managing 4.0 due to error correction)
clarity edit: errors on one system (PCI-E) caused a related piece of hardware to reset with it (USB 3) - this could be similar, with something resetting invisibly causing super fast stutter

this could be similar... it could be the RAM itself, it could be a barely managed error, or it could be a coinkydink (such as changing a riser cable and not thinking it related)

 

[tabascosauz]

temp related could make sense, if the RAM literally freezes/pauses for a moment, then wakes up literally a second later - one brief bit of stutter and its cool

whys it erratic? cause that one ram module simply may not be the one in use, the next time

The only thing I can think of is that it is some sort of bizarre never-before-seen "cold bug" on Hynix. If it chose to manifest, it would only be in the first half of the first match of the day. If I saw it and immediately quit and restarted, it would continue as soon as I came back, and would persist as long as I didn't suffer through the always initial 5-10 minutes of gameplay before it went away.

As you know RAM always starts out cold in the high 20s or low 30s, then slowly makes its own heat/roasted by the CPU/roasted by the GPU. MW is not a light game, always hits both CPU and GPU hard (even if framerate limited/DLSS/CPU or GPU bound). That said, either no other game suffered the same way or was never obvious enough to notice.

But MW isn't that much of a RAM heater, the CJR @ 1.38V was like 44C max. Almost the same temp as the B-die @ 1.5V, 43-44C. It's not nearly enough to trip B-die, which is by FAR the most temp sensitive of any IC.

For a long time esp. on the cursed 3700X I thought it was an IF issue. But it seems like a red herring in the end. I did have the random USB disconnect sounds from Aug 2019 all the way up till I switched to the Unify-X last month, but that's separate.

Hm, that is really interesting. I've been seeing some ... possibly similar weirdness on my system, in particular packet loss and latency variation warnings in Rocket League (it's the only online game I play with some regularity). Seems to happen regardless of the internet connection quality. This is with my 3200c14 B-die though (running at DOCP, still haven't bothered to OC it any). I might stick my old TridentZ kit in here just to see if anything changes.

As part of troubleshooting this, I've discovered that my system seems to get some pretty bad DPC latency issues after one or more sleep cycles, particularly when left asleep overnight. After a reboot everything is fine though. Any chance you could run LatencyMon with your previous kit after a sleep cycle or two to see if there is any relation? (I've been in contact with ASRock about this, and they say they can't reproduce it.) My RAM never exceeds ~45°C though.

I tried running LatencyMon a lot. It would always pick up on the nuanced micro-stuttering due to multitasking/monitoring software/other GPU usage while in game, but never on the extremely obvious frameskipping. Not once. LatencyMon wouldn't even register a small spike, just looked completely normal though I was literally missing the first 0.1 sec of a slide or inspect animation.

On the topic of the other microstuttering, moving to dual rank B-die was a CRAZY difference. The perceived reduction in input lag was like 60Hz>165Hz all over again. And not even the slightest hint of stuttering, I never realized how rough it used to be until I fired up the game on 3600CL14 and it was BUTTERY smooth. I could entertain the idea of placebo if it was at night or drinking (tired, less alert, everything feels "smooth"), but this is in the middle of the day.

Both are dual rank kits, so it's not the single rank min FPS issue. 3600CL16 CJR to 3600CL14/3800CL14 B-die.

Although, I'm debating going back to 3600CL14 because 3800CL14 doesn't actually improve anything in any game, lol. Saves me 80mV in VDIMM, 75mV in VSOC, 50mV VDDP, and a whopping 110mV on VDDG

 

[Valantar]

I tried running LatencyMon a lot. It would always pick up on the nuanced micro-stuttering due to multitasking/monitoring software/other GPU usage while in game, but never on the extremely obvious frameskipping. Not once. LatencyMon wouldn't even register a small spike, just looked completely normal though I was literally missing the first 0.1 sec of a slide or inspect animation.

On the topic of the other microstuttering, moving to dual rank B-die was a CRAZY difference. The perceived reduction in input lag was like 60Hz>165Hz all over again. And not even the slightest hint of stuttering, I never realized how rough it used to be until I fired up the game on 3600CL14 and it was BUTTERY smooth. I could entertain the idea of placebo if it was at night or drinking (tired, less alert, everything feels "smooth"), but this is in the middle of the day.

Both are dual rank kits, so it's not the single rank min FPS issue. 3600CL16 CJR to 3600CL14/3800CL14 B-die.

Although, I'm debating going back to 3600CL14 because 3800CL14 doesn't actually improve anything in any game, lol. Saves me 80mV in VDIMM

Hm, I've never watched LatencyMon while playing (only got a second monitor a few weeks ago, and I've taken a break in troubleshooting this for a while). Definitely worth a try! I'm getting DPC latency spikes in the 2500+ range just sitting on the desktop in web browsers and stuff though. And my kit is dual rank B-die, unless I have no idea how to read a Taiphoon Burner readout.

That being said, I feel like I'm doing this RAM a disservice by not tuning it some

 

[Valantar]

There was that thingy with ryzen and the WHEA errors causing the USB ports to reset, because of errors on the PCI-E bus (usually caused by PCI-E 3.0 risers mostly managing 4.0 due to error correction)
clarity edit: errors on one system (PCI-E) caused a related piece of hardware to reset with it (USB 3) - this could be similar, with something resetting invisibly causing super fast stutter

this could be similar... it could be the RAM itself, it could be a barely managed error, or it could be a coinkydink (such as changing a riser cable and not thinking it related)

Might be. I'm not seeing any errors in Event Viewer though.

 

[tabascosauz]

2 years of DDR4, Ryzen, and horrible AGESA experiences

 

[tabascosauz]

Here's an expensive lesson in

if
it
ain't
broke,
DON'T
FIX
IT​

When I bought my Cerberus X, I made the questionable decision to get a windowed panel, but the Cerberus X's SFX internal setup makes for a limited choice of viable coolers, let alone those that work thermally with a windowed panel. The windowed panel has been on the back all this time (offering a GREAT glimpse of the back of the socket ), and I wanted to put it on the front to show off.

So I wanted a cooler that wouldn't absolutely suffocate with the window on. Ergo, fixing what ain't broke.

• NH-U9S: one of my favourite coolers but certainly not better than the C14S, and 2 x NF-A9 are kinda whiny sounding
• Any of Thermalright "Mini" 120mm coolers: not enough heatsink mass, not available
• NH-L12S: too weak, not enough heatpipes, single-fan U9S-level performance
• Dark Rock TF: horrible RAM clearance, weird compatibility with non-BQ fans
• Fuma Rev.B: discontinued, barely fits with vented panel, cannot use windowed panel
• Fuma 2: does not fit
• NH-C14S: big chungus, clearance issues in push-pull, anything except vented panel murders performance
• Thermalright Silver Arrow 130: literally designed to cut your hands, weird fan size, 600g
• Thermalright Silver Soul 135: not designed to cut your hands, also weird fan size, 590g

Knowing that, doesn't the SS135 sound like the perfect cooler? Welllllllllllll.....................................................

Reminds me of that one thread arguing over the merits of downdraft coolers............................well, it's on full display right here

Needless to say, I went back to the C14S and the windowed panel went back on the backside of the case where I can't and don't want to see it. I guess it was a little much to ask of a ~740g cooler when it's going up against a ~1400g cooler.

And I'm sending the Corsair Vengeance RAM cooler right back as well. It's not terrible, but it doesn't cool as well as the C14S (imagine, a CPU cooler that does a better job at cooling RAM than a RAM cooler).

Oh, and the rough coldplate on the Thermalright also scratched my CPU. I might use this cooler as lapping practice since I can't really say I give a shit about it anymore. Once I get the coldplate fixed, might do a head-to-head with the U9S on some old hardware. I have a feeling the U9S might win with 2 fans, seeing the performance the SS135 has offered up so far, but who knows what it'll do with 2 x 120mm.

Remember the motto: if it ain't broke..........

 

[Mussels]

Are there better fans than the NF-A9 you can use instead for the NH-U9S?

be quiet have some stupidly quiet fans, p[retty sure they have a 92mm model?

 

[freeagent]

Sorry to hear brother, was hoping the cooler would have worked out for you. I won’t be recommending the brand anymore. It’s been good for me, but my experience isn’t everyone else’s

Sorry about your IHS too

Mine has a pit in it too, not sure how it got there

 

[QuietBob]

Oh, and the rough coldplate on the Thermalright also scratched my CPU.

How did that happen? It's difficult for me to even imagine how the cooler would do so much scratching on its own. What paste did you use?

 

[tabascosauz]

Are there better fans than the NF-A9 you can use instead for the NH-U9S?

be quiet have some stupidly quiet fans, p[retty sure they have a 92mm model?

BQ's fans are the quietest, definitely quieter than A9 but the U9S needs all the airflow since 92mm fans only push so much air. A9 pushes 46cfm, Silent Wings and Pure Wings push about 35cfm.

I was just speculating, I don't have both the SS135 and U9S to test at the same time, need to use at least one on another PC. I always use the C14S in my main.

Sorry to hear brother, was hoping the cooler would have worked out for you. I won’t be recommending the brand anymore. It’s been good for me, but my experience isn’t everyone else’s

Sorry about your IHS too

Mine has a pit in it too, not sure how it got there

I don't mind larger scratches as every one of my coolers gets some eventually, usually around the outline of the IHS from the mounting pressure. It's the pits/protrusions that are concerning, I don't think I've ever bought a cooler in my life that damaged a CPU.

Thermalright is aight, it's just not for me. I'd still buy a FC140 if I had a big case. But they really have to step up their fit and finish and QC.

How did that happen? It's difficult for me to even imagine how the cooler would do so much scratching on its own. What paste did you use?

It was mostly down to my stupidity, I already saw the pitting out of the box but it felt minor enough that I didn't think much of it. What it looks like is how it came, it had the protective film still on so I assume this is how it left the factory.

Not surprised from the rest of the fit and finish.

I just use NT-H2 these days. H1, H2 and MX4 have never scratched a cooler.

 

[QuietBob]

What it looks like is how it came, it had the protective film still on so I assume this is how it left the factory.

Oh man, if that's the case Thermalright really need to get their act together. One has to wonder whether it was inept workmanship or cursory QC? Regardless, sorry to hear about your mishap. My experience with their coolers has been very positive. I have four of their big Machos, and all of them perform admirably in standard cases. Perhaps they can't realize their full potential in a small one?

 

[freeagent]

It’s strange, my fc140 is mint, fit and finish was like my D14.. tight.. nice base to mount.

 

[The red spirit]

BQ's fans are the quietest, definitely quieter than A9 but the U9S needs all the airflow since 92mm fans only push so much air. A9 pushes 46cfm, Silent Wings and Pure Wings push about 35cfm.

I highly doubt that. I have read many reviews at TPU and final conclusion is that fans are all almost the same at cfm/noise ratio. There's is a difference at which rpm cfm and noise values are achieved between fans, but cfm/noise varies only very little. The only fan that managed to truly break away at high rpms was NF-A12, but that's it. Most fans perform the same as generic 7/9 bladers. However there are some experimental designs that failed to match even that basic performance level. In fact there are more failures to reach generic fan performance level than successes. BQ seems like a nice brand, but I don't think that their fans are any special. They often aren't going really low in RPMs. You want something that goes really low, Scythe makes that. Cooling performance isn't great, but they are quiet. Some of their PWM fans go as low at 400 RPM. Now that's impressive, but probably not viable in SFF PC.

 

[P4-630]

Some of their PWM fans go as low at 400 RPM.

If you can't control the speed of your fans by BIOS or software, with noctua you have an extra option

The NF-A12x25 PWM minimal rotation speed is 450rpm.

NF-A12x25 PWM

The NF-A12x25 is a highly optimised next-generation 120mm fan that integrates Noctua’s latest innovations in aerodynamic engineering in order to achieve an unprecedented level of quiet cooling performance. It takes state-of-the-art technologies such as the AAO (Advanced Acoustic Optimisation)...

noctua.at

 

[Valantar]

Be Quiet! only has the Pure Wings 2 in 92mm, and those are rated at 56m³/h vs. 79m³/h for the NF-A9 (at 1900rpm vs 2000 for the Noctua). Of course ratings and real-world performance are two very different things (especially given that there AFAIK is no actual standard for fan ratings), but that's a big difference. I doubt there are better 92x25mm fans on the market than the Noctuas.

 

[Chrispy_]

Be Quiet! only has the Pure Wings 2 in 92mm, and those are rated at 56m³/h vs. 79m³/h for the NF-A9 (at 1900rpm vs 2000 for the Noctua).

Is that because I'm hoarding them all?

 

[Chrispy_]

I highly doubt that. I have read many reviews at TPU and final conclusion is that fans are all almost the same at cfm/noise ratio. There's is a difference at which rpm cfm and noise values are achieved between fans, but cfm/noise varies only very little. The only fan that managed to truly break away at high rpms was NF-A12, but that's it. Most fans perform the same as generic 7/9 bladers. However there are some experimental designs that failed to match even that basic performance level. In fact there are more failures to reach generic fan performance level than successes. BQ seems like a nice brand, but I don't think that their fans are any special. They often aren't going really low in RPMs. You want something that goes really low, Scythe makes that. Cooling performance isn't great, but they are quiet. Some of their PWM fans go as low at 400 RPM. Now that's impressive, but probably not viable in SFF PC.

Currently running those PW2 fans at about 600rpm on my two radiators and front three intakes.
Granted, I don't bother overclocking either the 5800X or 3060Ti so peak load is only ~300W in synthetic combined loads and real-world is more like 250W.

It does depend how quiet your noise floor is though, because even though those fans will start and run at about 500rpm, they're quieter than my air-source heat extractors two rooms away as long as they're running below about 750rpm under the desk.

 

[Valantar]

Is that because I'm hoarding them all?

View attachment 216891
View attachment 216897

... those are 120mm? We were talking about 92mm fans. I think you interpreted my sentence as "the pure wings 2 series only comes in 92mm" and not "they only have 92mm fans in the pure wings 2 series", which was the intended meaning. Given that people were discussing possible alternative fans for the NH-U9S I thought that much would be clear.

 

[Chrispy_]

Oh, you Mean "BeQuiet's only 92mm fans are in the pure wings 2 series?

I skipped a few posts, admittedly thought we were looking at Tabasco's shortlist of CPU coolers still.

U9S with decently fast fans will be fine, I have eight 1950X with NH-U9 TR4-SP3 on them, it's surprising how well that thing cools even at modest fan RPMs.

 

[Mussels]

Guys, it wasnt about the performance: it was that he didn't like the noctuas noise profile

So if you gotta drop a noctua to 450rpm cause you hate the sound but can run a BQ at 1500RPM cause it doesn't vibrate the air in a way that tickles your ear hairs, you're getting more performance per tickle and you're happier.

 

[Valantar]

Guys, it wasnt about the performance: it was that he didn't like the noctuas noise profile

So if you gotta drop a noctua to 450rpm cause you hate the sound but can run a BQ at 1500RPM cause it doesn't vibrate the air in a way that tickles your ear hairs, you're getting more performance per tickle and you're happier.

That's true. Though then the question becomes if the Noctuas wouldn't sound equally pleasant at a lower RPM. They might not, but the impulse to not leav performance on the table can be very strong too.

 

[Chrispy_]

My experience of Noctua fans is that they're significantly quieter than many other brands at their full speed but not necessarily optimised for very low rpms.

BeQuiet's fans do seem to be excellent performers - both acoustically and airflow - at very low speeds. I wouldn't choose a BeQuiet over a Noctua for a high-performance solution requiring >1000rpm but for a use-case where the motor noise is more significant than the airflow noise, even their cheaper Pure Wings 2 are my preference over Noctua.

 

[Jdogdarkness]

Hmm. So i got my new 5900x and the problems persist. My write speeds are always messed up. Across all caches the read/ write/ copy ALWAYS vary across l2 and l3. For the l1 cache it very often. From what i understand a variation in over 20 gb/s shows some sort of instability. This problem is even on stock. I cannot figure it out i have replaced everything. Last thing i have is i replaced my PSU with the same PSU model. Maybe 5900x doesn't play nicely with corsair rm850x? But then again my laptop is starting to come out of sync too. (4800h). Could i have like bad electricity that is frying my hardware?