*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

I haven't done an OS reset for a long time. I made new Macrium Reflect images after the clean install, before a ton of software crap was installed. The Macrium images I had before ended up being corrupted due to a bug in Macrium Reflect, so I could not restore them. I also clean installed W10. I had been running only W7 for several weeks because something got buggered up with my W10 installation and I was getting 5 or 6 random BSOD ever day, all referencing ntoskernl.exe, sfc /scannow had corrupted files that could not be repaired, etc. When I wiped that 960 Pro a few weeks ago all of those problems vanished.

 

OK, added the auxiliary pump and reservoir to the GPU and added QD fittings and 0.5 in ID hose to the CPU. Also installed a 5.25 inch bay dashboard with USB ports and front panel audio. The red color is the WaterWetter I added to the distilled water. Not my favorite, but it only comes in red and it is not designed to be an aesthetic additive. Would have preferred clear.

 

Hey Mr Fox, you don't have to explain away your pink unicorn water, you can have whatever colour you want! ;-) But, on a serious note, does the water wetter make much difference to temps? Last time I heard of water wetter it was being used in modified turbo charged cars.

 

My temps do seem to be a little better, but it is not possible to say if it is the additive or the physical changes to both loops. The variance among the CPU core temps is less now, which may have something to do with how the water contacts and flows over the tiny fins in the CPU block using the WaterWetter (which is by design). I noticed that more than anything else. I will know more in a few days, after I have more time for observation. It seemed like a good alternative to the overpriced additives sold for PC cooling, and I saw where others mentioned using it at Overclock.net.

The compelling aspect of using the WaterWetter was the prospect of the additive improving temps, whereas most additives go the opposite direction compared to using distilled water with no additives.

I do find the color annoying and may end of flushing it for that reason alone. I added a drain hose and valve to the CPU loop, so that will be really simple.

I do not care for the GPU pump/reservoir, but it was relatively inexpensive and works well. The diameter is not as large as the CPU reservoir and I do not like that the reservoir and pump connect with a male-male G1/4 fitting with an air gap between them. The pump has threads to screw the reservoir into the body of the pump, but the reservoir only has threads on the inside of the tube. I would like it better if the two parts fit together by the reservoir screwing directly to the pump without the fitting and small gap between them. I am probably being way too picky about that, because it is not really all that noticeable. But, I am good at being nit-picky, LOL. If it had some kind of filler or spacer to fill the air gap that would be nicer. I also do not care for the snap-on clamp style. It does not take a lot of effort to unsnap the tube from those brackets. I like the solid mounting of the XSPC setup a lot better, so I will most likely buy a matching one later on for almost 3 times the cost.

My only real criticism of the Praxis WetBench after having more time with it is the drive bracket and center tray slots, and the thumb screws. The slots in the center tray for attaching drives and the optional 5.25 inch bay bracket are not thoughtfully spaced and the thumb screws are made of plastic. Small complaints in the grand scheme of things, but considering the cost of the bench, those parts should have been given more thought and revised at some point. I will give Praxis some feedback on that. It is not enough of a big deal that I would not buy another one, but enough to be irritating while I was figuring out how to work around the poorly arranged slots in the center tray. The plastic thumb screws are really chintzy and I don't get that at all. Forgiveable, but worthy of mention. Brother @Johnksss, what say you?

I ended up mounting the two 1TB Samsung SSDs on edge, perpendicular to the center tray. That worked well. Due to the spacing of the slots, laying them flat side by side did not fit well with the spacing of the SATA power connectors on my PSU cable. The distance between the connectors on the cable was not enough to reach the drive spacing with the drives laying flat and I would have needed to use two cables with unused connectors between each drive. To be fair, I had to do that with both previous cases as well, but the extra mess of having to use two cables with unused connectors in the middle was essentially concealed behind the mobo tray. I found that equally annoying with the previous cases as well. Ideally the PSU cable would have 6 to 8 inches of space between the SATA power connections.

 

Well brother @Mr. Fox , Not really sure if you were just setting your's up differently, but the drives go on top of the middle section. That tray under neath is for 5 2.5 drives or accessories. Normally.

Just some ideas....

And the cables can be tie up under the tray to not show. And facing the drives back to back should take care of the connectors not long enough to reach.

This one involves spending a little money.


That would hide all your drives. I actually have a one like that for 4 2.5 and 4 3.5 drives.


As to the plastic thumb screws. Metal would have been better, but the nylon ones don't bother me. I guess they used them to help keep scratching the case up. Speculation of course.

 

I had the drives like that at first. It worked fine until I tried to add.the 5.25 inch bracket and that caused a conflict, and it also requires using two SATA power cables and extra bulk without using all of the ports because the drives on right are too far from those on the left. Maybe I am just nit-picking too much. Perhaps I should look into having a custom SATA power cable made for my PSU with more space between the ports and only the number of ports I need. A round sleeved cable would also be better than flat.

 

I'm thinking you would benefit more from sleeved cabling/combs/material cable ties and a sleeved cable extension. (if need be)

Here is mine


Of course this is not the finished product. Just pulled it apart to show


No way to get rid of bulk using the factory cables.
But again, these are only suggestions.

 

Yeah, you're probably right. The factory cables are also way too rigid and unwieldy. I have hated them since day one. No point in having extensions because you have those nasty harder-to-manage factory cables to deal with, LOL.

I see you have the brackets flipped the other way, with the flanges at the top and bottom facing away from each other. That could be part of my issue with the slots lining up. I have the flanges facing toward each other and it's not the correct spacing for 5.25 inch drive width. I am going to check that out later today and see if it helps.

I am going to send you a PM about something else here in a bit.

 

PM?
Just call. LOL

 

I ended up managing OK with the cables I had. I found some SATA 6GB/s cables with 90° ends on them that I had purchased last year in my stuff and that allowed me to place the two 1TB SSDs facing one another on the center tray. I kept the 2TB Barracuda HDD on the 5.25 inch bracket. Flipping them around with the flanges facing away from each other allowed the slots to line up, but can only attach it with two thumb screws if I want the bay near the front of the bench because the slots in the center tray do not come forward far enough to install four thumb screws. Cables are routed neatly now.

Spoiler: Pics...

Incoming...

Spoiler: What's incoming?

For the discrete GPU water loop...


If anyone wants to purchase the EVGA Hybrid setup with the Thermaltake 360 RGB radiator, fans, controller, etc. let me know. I will leave it set up with the quick-disconnect fittings. (The new pump and reservoir will not be included, but they are not needed with the AIO pump. The View 71 TG RGB is also on Craigslist, but available to anyone that wants it. I will include the vertical GPU bracket and PCI-e riser cable. I can put both in the NBR Marketplace.)

 

Note the temps and power draw is 837W running with 8K Optimized preset, LOL. No way this 1kW PSU will be able to handle two modded GPUs like this.

Unigine Superposition 1080P Extreme: http://hwbot.org/submission/3893625_


Unigine Superposition 8K Optimized: http://hwbot.org/submission/3893626_

 

we should rename this thread into smth like: "Johnksss & Mr. Fox Xtreme Desktop Mod & Bench Thread (Lurkers welcome!)"



not complaining here! just that compared to you two any other posts seem almost dull in comparison haha

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Ha, well if you're not buying any new hardware then you don't have anything to show when it comes to that aspect - they both upgrade often! But, from my point of view I find any posts interesting related to performance insights/news, case performance mods, etc, anything of that ilk - so it doesn't have to be just a place to show off new hardware.

 

Nice. Glad it all worked out!

Looking good! I did a run like a day ago and got a 6505. Today after the new Bitspower water block, it went up a little.

 

I ran it again. Best I could pull was 630W to 634W.
1.46V Memory
1.41V GPU Mem
1.17V GPU Core
1.385V CPU

 

I wonder if the Kill-A-Watt meter is not polling fast enough to see the peaks? Either that or my CyberPower UPS meter is not accurate. My Kill-A-Watt doesn't match the power draw in a more strenuous GPU test like 8K Optimized (saw 900W with the CPU at 5.4GHz), but in the 1080P Extreme it is much closer. I would think your i9 @ 5.0GHz and K|NGP|N 1080 Ti would pull at least 900W peak under severe GPU load. That 634W seems too low to me for what you did with those benchmarks, but maybe not.

Here is a video showing the difference between the two power meters. Even Fire Strike was pushing just over 800W peak.



Superposition 1080P Extreme - http://hwbot.org/submission/3894185_

Superposition 8K Optimized - http://hwbot.org/submission/3894188_

https://www.3dmark.com/fs/15903063

 

Just want to share a few Results

Desktop

my daily driver setting:

https://www.3dmark.com/fs/14946883

i7 5960X @4500 (actually 4.625ghz)
32GB DDR4 3000
EVGA GTX 1080Ti SLI @2025mhz
Corsair AX1500i
EVGA X99 FTW
Asus PG348Q

Cooling:
CPU: Aquacomputer Cuplex Cryos
GPU: Heatkiller IV
Water: Aquacomputer Airplex CU, MO-RA 3 420

Notebook
Aw17 R5, 19 8950HK, GTX1080, LM repaste

Cinebench @ 4.7ghz


3D Mark FS

https://www.3dmark.com/fs/15887876

GTX1080@2025mhz
i9 @ 4700

 

Not really sure.
Mine is pretty accurate after I bought a new Kill-O-Watt meter. The first one I had was like 250W off half the time.

I just ran a 1080 Extreme at 2151/6497 and got a slightly higher score than you. Max watts was just under 300W for gpu.
Yet, on firestrike the SC cruises past my card.

 

It's too bad Superposition does not give us a separate rating or score for the physics part of that benchmark so we could more easily understand how your monster CPU plays into the overall score. Only a guess here, but knowing how Fire Strike works, more credit is given for GPU performance than physics and my assumption is that Superposition doesn't discount the value of that amazing CPU like Fire Strike does. There is no way I should have beat your Fire Strike score with an 8700K with our GPUs overclocked so similarly. I think Fire Strike isn't giving us a clear picture of total system performance and focusing mostly on the GPU in the overall score. But, we already knew that and have known it about Fire Strike for a long time, LOL.

 

I don't think CPU does anything for Superposition, last time I checked it hardly uses any CPU resources, I'll run it again & check. Actually, I can't, I did a fresh install recently and don't have Superposition installed, but pretty sure it barely taxes a CPU from my memories.

 

Well, my score did go up when I bumped from 5.2 to 5.4GHz. And, the extra 10 cores/20 threads with the 7980XE should definitely come into play somewhere. To your point, there is not a huge difference between the Superposition scores. Not as huge as one would expect with a 300% increase in core/thread count. Since my GPU was overclocked a little more, I think that is why it is favored in Fire Strike (because of how 3DMark designed Fire Strike) and I believe the extra bandwidth/cores/threads of the i9 are what pushes it ahead in Superposition.

 

I just remember it nowhere near taxing my CPU, not sure if it even got to 25% usage, if I recall. (You'd think it would be able to use more than 1 thread of a CPU.)

 

When I say beat. I was referring to your gpu score. And with Firestrike, cpu does matter.
https://www.3dmark.com/compare/fs/15651270/fs/15284270
With Superposition, I think it's more of a single core that helps the most and not how many you actually have.

I think you are correct in this assumption. It uses a small a mount because it's not taxing all the cores of the cpu.

As well it should, but not because of how many cores you have, but more so how high your single core status could hold the 5.4 multiplier.

And along that line of thinking.....What would be this guys excuse?
http://hwbot.org/submission/3760521..._8k_optimized_geforce_gtx_1080_ti_5128_points
I'm pretty sure he used the LOD mod to get that score.

Edit:
Never mind, he ran it in custom and turned off motion blur and low textures.

 

Seems like that needs to be reported for non-compliance. I just checked the rules and it seems no validation link is required for Superposition (which is unfortunate). Did you find the link where he submitted it to Unigine? I see those disabled in his screen shot.

Superposition rules say the following is not allowed:

• Altering benchmark files or the rendering
• Any software or human interaction altering the perceived speed of the benchmark program, tricking it to believe it ran faster
• Lucid Virtu MVP
• Mipmap
• Change benchmark settings

 

More photos... no time for benching right now, but it looks good. Had to elevate the pump for the GPU.

Spoiler: Open Sesame

 

Aren't they the same as you other photos a few pages back, what's the difference? Was it the elevated pump that's different, couldn't see difference as not sure what looking at.

 

Exactly what I was thinking. Considering his is obvious. some of those others, not so much unless you run the benchmark with the same hardware and cooling.

New GPU Hydro Copper water block my friend.

@Mr. Fox The hydro is looking real sexy!

 

And, larger XSPC radiator (thicker). See links in signature for the old stuff.

Looks nicer, but I'm not sure it improved my cooling enough to justify the added cost. Maybe 3-5°C better, but haven't had time for extensive testing yet. Core didn't really need to be any cooler, but the GPU power delivery system may have benefited more than it seems from my initial impression.

Thank you, Brother John. I think it looks great as well.

 

Well, I finally had time for the "after" testing to compare with the "before" testing. It appears cooling the power delivery system has increased the GPU core temperature just a bit... not enough to matter... but there is a pretty big drop in the temps of the memory (10-15°C cooler) and power delivery system (20°C cooler) with the Hydro Copper block. All things considered, I'd call it a win. These results are after two back-to-back 20-loop Fire Strike Ultra stress tests on each cooling system.

BEFORE

AFTER

@bloodhawk @Johnksss

 

What's going on around here? I'm away for a while and everything changes. I was just about to email you to tell you to check out a cool bench rig I found from PrimoChill. I guess you guys already found it. That thing looks mint. Did @Johnksss build an open air rig too?

 

Thanks. It does look pretty awesome. Yup, Brother John did it, too. He got his first. I did not think they were available any more, but PrimoChill must have had their stock replenished. Brother John always has open-air rigs, but got his Praxis WetBench on.

 

WOW! This is insane! I read somewhere that the i9-7980XE only over-clocks to around 4.5ghz. I guess they were wrong.

 

"They" usually are wrong... about 85-90% of the time.

But, nothing overclocks well unless (a) it's a nicely binned sample, and (b) the user knows what they are doing.

 

Three loops of Heaven, followed by benchmark (loop #4).

 

Might have to do with the voltage being higher....maybe.
And it's looking like a win to me. It just looks better over all. And no performance loss either. It is matching the gpu numbers we put up in April with chiller on cpu.

Yep. I indeed went out and bought one. Gotta say, I like it a lot. The only real thing I don't like is there is no easy way to run it flat for LN2/Dice benching.

pretty much depends on the cooling/and user as @Mr. Fox pointed out. Although this thing will do 4.6 - 4.7 without really doing to much of anything.

 

1999.5Mhz? What happened to the 2200Mhz I'm used to seeing your GPU at? Or the GPUz screenshot suggests it might just drop to 1999.5Mhz when not loaded, but in some of your older more historical screenshots it just looked locked at 2200Mhz.

 

I ran a 3DMark 11 at 2200. Did you see it?

Overclocking doesn't really change the temps that much, but I did the stress tests stock (yes, 1999 is stock boost on this GPU) because that was a WHOLE LOT more sustained load for a much longer period of time running 40 loops of Fire Strike Ultra. I did not want to over-stress the GPU by being abusive to it, just wanted to verify before and after temps. I don't mind working the CPU really hard because it is more durable and costs about 60% less than a 1080 Ti, LOL.

 

Oh right, cool, I didn't realise that your 2200Mhz setting was on the borderline dangerous level - I mean I know 1.2V is definitely not standard (!), but figured you'd be the kind to run all your games & everything at 2200Mhz, especially given the low temperatures. But yes, 1080ti is a lot of dough, and gaming probably doesn't need 2200Mhz! (oh yes, just noticed you ran the 3dMark11 at 2200mhz).

 

I don't know that it is dangerous per se, but these stupid GPUs are so expensive I reserve the overclocking mainly for benching and short burst of fun. I generally play games with the GPU stock. Overclocking provides minimal benefit in games unless the GPU is already too weak. I did do a couple of YouTube videos with the GPU overclocked just for show, but it adds maybe 5-15 FPS to the already above 100-150 FPS, so no real world benefit I can identify.

 

That always depends on the personal requirements settings, AA, resolution, DSR, desired framerate ect.
With the right settings my 1080Ti SLi can limit at 1920x1080 with one-digit frame rates or can get bored in 3440x1440

@Mr. Fox is your 2200mhz setting gamestable?

 

I guess so if you crazy with DSR & AA! Although 2200Mhz in comparison to 2000Mhz stock is only a maximum of 10% increased performance (best case, because doesn't scale perfectly) - so this would mean going from 30fps to 33fps or 60 to 66fps, 120 to 132fps (you get the idea!). That's not really a massive difference any way you look at it - unless you don't have G-sync and it just tips you over the V-sync limit.

 

If you can push your min fps above your gsync/vsync Limit then even 10% oc is worth it. but Yeah usualy you dont need to OC for normal gaming.

 

Hmm, this could be interesting - my CPU is no longer completely stable at the same voltage & frequency that it once was. Previously 4.7Ghz @1.4V and 4.6Ghz @1.344V were stable in Prime95 (non-AVX v 26.6) and OCCT. Now 4.7Ghz is not stable on Prime95, fails after about 20mins, while 4.6Ghz is stable on Prime95 for an hour+ but not stable anymore on OCCT (fail within 5mins). I'm wondering if my CPU has degraded. I've been running at 4.7Ghz, and recently playing a lot of Battlefield 1 that really stresses the CPU (not been crashing on that), so perhaps the extra strain has slowly degraded the CPU over the last year. The only changes I've made: new BIOS with new microcode, so perhaps that plays a part in the stability, plus it's also 28 degC here at the moment when usually it's about 20-22 degC (room temperatures). My CPU is pretty much always below 60 degC (apart from in this barmy Summer right now), so I'm surprised that I might be seeing degradation. Gonna run it at 4.6Ghz at the same 1.344V to be safe, that's at least Prime stable, although lost it's OCCT stability.

A few months ago I did have some stability concerns when I saw crashing for the first time when running Firestrike loop combined with Prime95 - and I found out it was due to overheating VRMs on the motherboard (saw thermal throttling in HWInfo). However that was just due to the raised temperatures in the case due to the GPU load and CPU load, also HWInfo not showing up thermal throttling now with just CPU load. Perhaps the VRM's are still not happy in this hot weather or maybe it's CPU degradation, or maybe it's the new BIOS with the new microcode - I can't be sure. But, I think I'll tone it down to 4.6Ghz at the same 1.344V to avoid worsening the issue (CPU degredation, or VRM woes). Perhaps you folks out there be careful with CPU voltage in the mean time - I had had mine at 1.4V for over a year now.

(I might try turning it back up to 4.7Ghz in the Winter to see if it's magically stable).

 

Have you tried with the older/previous Microcode patches from Windows update? You are on the latest (I mean it's 0xC6) now? You can even try older one with the Ucode tweaks (Linux* Processor Microcode Data File). As well with meltdown/spectre protection disabled.

 

2000 mhz is not stock. stock is 1556 mhz.

 

Well yeah, stock for the 3rd party card, they come pre-overclocked. Ha, it ain't really overclocked unless you've done it yourself!

 

I'm on the latest C6 microcode that came with an updated BIOS. I could probably use Windows Update to 'downgrade' to C2, which is what I was on before - and then I could test stability again. I might do that sometime over next couple of days & I'll report back.

 

I didn't wait, I tested it just now! I used Windows update catalog to download C2 microcode - it still fails OCCT 4.6Ghz after about 5 mins. No thermal alerts in HWinfo. CPU at about 62 degC on hottest core. With Spectre protection disabled using InSpectre still the same failure of OCCT. Either something inherent in the new BIOS caused the instability, or the current hot weather, or maybe some dodgy and slowly failing VRM on my motherboard (because they were overheating a few months ago with a combined Prime + Firestrike load), or CPU degradation.

 

You compare the results with same OCCT version? There have been a few changes in the test software. And you have same max temp as before?

 

Yes. Temps are fine. Just haven't really seen any benefit from it.

My monitor is 2560x1440 165Hz. It has G-Stink, but I have that and v-sync disabled in my NVIDIA Control Panel. I do not care for either one.

I am experiencing the same issues with stability after applying the latest ASUS BIOS v1602 released 07/10/2018 with new micro-code. My RAM is unstable and the CPU takes more voltage. I went back to the original BIOS when the motherboard was first released trying to recover from the damage. If it remains unstable I will RMA the board. Too early to tell if the damage was undone by the old BIOS, but I am sort of apprehensive. I could not POST with all four sticks of RAM at XMP 4000 with the latest BIOS. Only bootable with two sticks at XMP 4000. I spent 4 hours running Memtest86 with one stick at a time in each slot. All were stable in all slots, stable with 2 sticks in either bank, but as soon as I put in 4 sticks I began having problems with Windows BOOTMGR and with Intel ME and checksum errors at POST.

I don't know if it is the new CPU micro-code or the ME that is buggered up. I have asked @Prema if he has any ideas.