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

Ah, now using HWInfo I can see the two drive temp's and that 2nd one gets higher. There was a run and I saw some minor throttling.

 

Then you can imagine how it will work in a thin flimsy Notebook workstation chassis(hard ssd use)

 

under atto, probably like junk. Normal access it should be fine, high end work well forget it.

 

Unless they're noisy? You don't want noisy if you've got a 'quiet' watercooled build!

 

Looking at the Temp2 of these drives apparently the controller runs quite hot, about 9c more so than the memory chips. With the HS as installed it seems to run even under normal usage pretty cool. Doing a Macrium back up to a SSD that the EVO 960 only got to 65c on the controller and normally seems to sit at 54c.

I've decided to try a heavier passive solution to see if there is a difference. So I have ordered; https://www.amazon.com/gp/product/B078X8XYJS/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1 .See if these really help or if they are a waist.

 

Well post some pics after you get one

 

Why not active cooled Heatsink when you first want try something new? $10 bucks more won't empty your wallets

 

I want to stay as quiet as possible. Actually it is 20 more as I need two. I am not a RGB fan but those may be my next purchase.

 

Albeit minimal there is air flow at the M2's

 

I increased the default of my GPU fans to a minimum of 40% and it lowered the EVO's by 4c both on the memory chips and controllers.

 

I just ordered one of the passive models from Newegg. I decided I didnt like the height of the active model. I may change my mind later, if @TANWare is going to get some, and posts some pics

https://www.newegg.com/Product/Product.aspx?Item=9SIAC8W6SF4126

 

I ordered the massive one from Amazon. It is huge too. One of the two did not have the bands or thermal pad so I have to return it and get another. Guess which has the new HS?

Edit; both now have the warranty sticker removed.

Edit 2; First run of ATTO only got to 56c on the controller and 39c on the memory, compared to 68c on the memory before at first run. Second run to 59c and 42c, third run at 60c and 43c, lastly the fourth run at 61c and 44c.

Well below throttling temps, if you can afford the room these are the ticket.



After the four ATTO runs

 

I should note just a single run of ATTO on the other, C:\, drive brought the controller to 69c and memory to 53c. I am not going to torture it further without a proper HS.

 

Wow, yes, they are tall! Way to maximise surface area! As well as increased heatsink weight to allow for soaking of more heat before temps start to rise.

 

The thing I like about yours, besides cost, is the height is not too intrusive. These I am going to be using are HUGE

 

Got a quick question. I bought a EVGA 1000 Watt Titanium T2 power supply but my VGA Cables that came with it are only 8X6 when the Ti requires 8X8 connection at the GPU.

Any idea what I should do here? Can I connect 2 separate VGA (8X8) cables to the GPU from 2 separate ports on the power supply or would that fry my GPU?

Thanks and sorry if this is a bonehead question

 

The JonnyGuru.com review says u get x4 single-8pin cables (6+2) + a pair of 8/6 types. Use the single 8pin pair. Had the full cable set been just an 8/6 design (as some PSU models provide), you'd still use two separate cables (lay / ziptie / velcro the resistive 6pin jumpers to the side), for best result

edit, per your photo, just use that pair of single 8pins


JonnyGuru advises to go by the Seasonic Chart above, your 1080Ti should run on two cables, don't sweat it that the cable pouch didn't have an 8/8 resistive jumper cable type. 8pins are rated for 150w / 6pins rated for 75w, thus an 8/6 was good up to 225w ... ... ... the reality is the load placed on the single 8pin modular connect at the PSU end of the cable, you don't want the jumper 8/6 8/8 to pull too much through it (when a connector fails / burns up / smokes, it is almost always the PSU 8pin header port from an 8/6 8/8 jumper pulling too many amps for too long, ie too much card for the cable ), so spread the load over two cables instead

 

Thats strange, I have an evga 850w power supply, and it has 8x8 connections.

 

Could just use 2 of these right?

 

Yeah I don't know what's up with that. I did buy it used so maybe they kept that cable. I have a 8+6 connector but no 8+8.

 

That is an 8x cord. They come with a 6 pin that has a 2 pin that slides into the socket with it. Those are exactly what I have, so I use two of those.

 

Yeah but what had me confused was that my old psu had this cable which is 6+2 plus 6+2 which i only had to use one cable to connect my gpu.



We all good now, she's running like a champ!

 

Best score so far @ 4.0 GHz;

 

Here is mine...My Disk is killing me. That's next on the list of upgrades. 5.0Ghz @ 1.35v

 

Finally got the second Heatsink on the C:\ drive;

 

http://oc-esports.io/#!/round/roadtopro_challenger_season4_division4_round1

 

http://hwbot.org/submission/3795393_

 

Time Spy: https://www.3dmark.com/spy/3363657# | http://hwbot.org/submission/3795780_

 

XTU Benchmark: http://hwbot.org/submission/3796204_



221.9W

 

That's a lot of Vcore you're using there for your CPU, 1.52V - do you just use 5.2Ghz for a few benches and then turn down the frequency & voltage for 24/7 general use? My research I did in relation to overclocking my 6700K suggested that 1.45V was the safe upper limit for 24/7 general use, with a lot of people chickening out around previous conventional wisdom of 1.3V. I'm using 1.4V for my 6700K, I'd consider using up to 1.45V but it doesn't improve my overclock.

 

It's not that high under load. You're looking at idle voltage, which can be very misleading. Load voltage is like 1.425V or less (can drop as low as 1.375V in Cinebench or wPrime 1024M). I run the CPU 24/7 like that, not just a few benches. I have it set in the BIOS for 52x6 with static voltage (which is never actually static) set to 1.475V. I can lower the voltage a little, but benchmark scores go down. Temps are fine, so I'm not worried about it. This is just an ordinary/average 8700K sample. Some can use 1.375V-1.400V for 52x6 with cherry picked nicely binned samples. As you know by now, I do not go with popular "wisdom" on voltage. I never have because I'm not worried about wearing out my CPU. If it gets voltage fatigued, I'll just get another one to replace it. I've overclocked and overvolted the crap out of all of my CPUs and that hasn't happened yet... 2920XM, 3920XM, 4930MX, 4930K, 4960X, 4790K, 6700K, 7700K and now 8700K have taken a lickin' and keep on tickin' LOL. Temps are good, so we're good to go. If you let the temps get bad, then you're courting disaster.

 

That's weird, the minimum voltage seen in that screenshot is 1.488V during your 9 min measurement period, which is not the load voltage of "1.425V or less" that you quoted? How come your idle voltage is lower at high loads in comparison to when idle, is that vdroop? Isn't it better to use some load line calibration so that the voltage stays the same for high loads as it does for low loads - at least that way you wouldn't have to pump more voltage through it at idle for it to maintain stability? I have loadline calibration set to Auto (think that's the only option in my motherboard), and loaded CPU Vcore is always 1.376V-1.400V depending on load (and I have 1.4V set in the BIOS as the target).

 

Well, 1.488V is what the sensor in the BIOS reads at idle and that is all XTU can see. XTU is broken and does not show any readings correctly and most of the controls do not work (or they are missing) with this motherboard and 8700K. Intel needs to release a newer version of XTU that functions correctly. Right now it is useless for anything except for the benchmark and setting multipliers and power limits. Cannot even adjust memory clocks with it (because the controls are missing). Plus, I have SVID disabled in the BIOS which kills all but the PECI voltage and watts sensors in all monitoring tools. With SVID disabled, DTS sensors are turned off. Notice how many things in XTU have bogus values. I have tinkered with LLC set to 1.0 and 0.01 and it does not seem to make any major difference. It does reduce vdroop, but does not eliminate it. From 6700K forward, every system I have owned seems to have "adaptive" behavior even when using static voltage... never stays locked 100% of the time on what I set the voltage to with any of the systems I have had since Sky Lake.

 

I'm not looking at XTU, I'm looking at your HWInfo64 Sensor readings - it says 1.52V max, 1.488V min? Are you saying they're bogus/invalid values there in HWInfo because you have SVID disabled - why disable those if they provide useful measurements?

 

Yes, disabling SVID throws all of those sensors off. They are somewhat close, but not accurate or reliable unless SVID is enabled. The Core VID sensors further up near the core clocks in HWiNFO64 also totally disappear when SVID is disabled and the CPU Package Power sensor shows a max of like 11W with SVID disabled. The measurements are useful, but I don't need to take measurements while I am benching because I already know the behavior of my system from watching with SVID enabled. Disabling SVID produces better results for overclocking. Disabling SVID turns off the CPU's ability to auto-regulate, which is essential at higher clock speeds. If you enable K-Boost, it does similar things to GPU sensors because it blocks communication with the firmware to stop dynamic behavior.

 

Ah right, so those values for VCore are bogus in HWInfo then, and you know how Vcore behaves from past monitoring when SVID was enabled, I see. But, you say CPU Package power shows stuck on 11W, yet in your HWinfo screenshot it shows 221W for CPU Power as a max figure (you actually highlighted it in the screenshot), or is CPU Package Power a different variable to that one?

 

No, that is a different sensor. I know it is hard to understand based on brief comments. Here are two examples. One is with SVID enabled, stock voltage and using ThrottleStop to set voltage and multipliers. Notice how crazy high the voltage goes at idle and how low it goes under load. Huge swing in voltage overshoot and vdroop (erratic both directions) using default Intel voltage control garbage. The other screen shot is with SVID disabled. Note the difference in sensors, missing Core VID and CPU Package Power shows 3.8W (normally maxes out around 11W). In the second screen shot with SVID disabled, the CPU Power max is almost the same as the CPU Package Power in the screen shot with SVID enabled. Yes to your question... these are different sensors, but I think they might represent the same measurement, although I am not 100% certain of that. The same value is set for the voltage in the BIOS, but the voltage behavior is more erratic with SVID enabled.

Hope that makes more sense. The bottom line here is software sensors are sometimes not 100% accurate. When I manually set values in the BIOS I have no reason to believe they are not being applied correctly. C-states and P-states probably cause sensor accuracy degradation, as well as erratic changes in power and voltage delivery. (I also disable C-states to improve stability.)

 

Thanks, I know what you mean now. I wonder why you get so much v-droop, I'm not sure why that is, I don't get that on mine and I've managed to push up to 140W Package Power - does it come down to motherboard or max power draw perhaps that determines if you get v-droop or not? On mine I have 1.400V set in BIOS, with a fixed frequency of 4.7Ghz, C-states enabled, and load line calibration on Auto. I did some short tests now to determine my own voltage fluctuation:

• wPrime1024 50% CPU(4 threads) usage results in 1.376 - 1.400V fluctuation
• wPrime1024 100% CPU usage results in 1.392V steady
• OCCT 100% CPU usage results in 1.392V-1.400V (mostly 1.392V steady)
• Prime95v287 100% CPU usage results in 1.344-1.400V (mostly 1.392V steady, 1.344V blip was when workers changed tests I think, it was still 1.392V steady during the long 136W sections) - see screenshot attached for Prime95 test.

I don't think I'm experiencing v-droop here, well maybe from 1.400V down to 1.392V if that's considered. I wonder why you have more? (I think I remember you having a top of the line motherboard too, so I would think motherboard would be main determiner for v-droop or not?)

 

Ummm, you have a desktop so you can measure from the board..

 

I guess I could, I don't have a multimeter. I have to rely on what HWInfo is telling me, (which I think Mr Fox has been doing too).

 

If you want the real then it will be a little more than you guess. That's why they have voltage points, because software can stories while the voltage points will tell the truth. It reads Voltage & VRM's just to name a few, but if your board is a basic one then you won't have them.
Then once you confirm that it matches with software, then you can say the software is correct.

 

Fair enough, but I think Mr Fox and myself are comparing software readings, that's what we're going on, understand your point though.

 

Fair enough.
I was only saying that because these are not laptops.

Side note: You can also use the motherboards software as well....If it came with it.

 

(yep, desktop sub forum)

 

Where they keep getting posted in the laptop section....

 

Using the software that ASUS provided, with static voltage I am not seeing the wild fluctuation observed in HWiNFO. The voltage monitoring read-outs do not fluctuate much at all in their software, so I think the third-party generic stuff is maybe in the ballpark, but not entirely accurate. If I run something like wPrime or Cinebench, the ASUS software doesn't move more than about .002-.003V compared to what HWiNFO shows.

 

Either that, or Asus want to paint a more rosy picture with their software, HWInfo might be more impartial. ;-) Ha, who knows! HWInfo doesn't show wild fluctuations in mine.

 

Yeah, right. Who knows.

The other side of the coin is I am not worried about it. I started to say I don't care, but that would not be entirely true or I would not monitor anything. I'm definitely not worried about it. I never spend any time thinking about it except when someone asks as you did.

 

Hummm...as desktop boards go. They pretty much rank number 1 in the world. So I'm guessing that that is not the case on their high end boards. That's why they come with voltage points.
HWinfo is really good, but it does not beat voltage points or true temp readings in no way, shape, fashion or form. That is a fact. Try using it under LN2/Dice/LNe/PhaseChange/-30 Water chiller and you will understand.

Now some of their returns departments...... Yeah, they need some serious work!!!

 

Well I was talking about the voltage fluctuation / vdroop in the context of it being bad for overlcocking because you end up having to run a really high voltage at idle in order for your high load overclock to be stable - that's not good for the CPU. But yeah, you know you don't mind if the CPU wears out in a couple of years - not that it will, but like you said you don't mind, we run these risks when we overclock beyond a certain point. And it's interesting to know why one board doesn't show much fluctuation and another board does - it's just an interesting learning point, you'd want to choose the one without fluctuation all things being equal.