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

Are you looking for a everyday stable OC? I know you like to push your memory to the limits, so a high memory OC with a high core OC can be very unstable.

 

the maximum reached is 4200cl17 18 37
10 hours ramtest and hci,cpu 5ghz

It should be noted that 1 hour and 2 hours of realbench with 1,210v pass occt.
But it turns out that to play BF5 I had to raise the voltage to 1,230


I don't know if I'm wrong but there are many things in the bios that barely influence the oc,

by the way right now I have it in stock but with a voltage 1,140v

 

The way I see it for CPU OC stability VCore is number one importance, then your load line calibration settings to make sure you're not getting vdroop at high CPU loads - ie you want your VCore to be 'the same' at both low and high CPU loads. For RAM OC stability you have RAM voltage, VCCSA voltage, VCCIO voltage, playing with the frequency and latencies. Also, the cooler you can keep your CPU (and RAM) then the more stable it's gonna be, in terms of you might be able to get a slightly higher stable OC if they are really cool vs if they are very toasty at say 90 degC.

 

stability test of about 4400 cl19

 

4400Mhz CL19?? Your screenshot is showing 3591Mhz at CL15. What am I missing here? (Also 4400Mhz CL19 would be slower than your previous 4200Mhz CL17 - worked out by dividing the frequency by the CAS latency to work out a rough indicator of comparative performance, so probably no use for you to try to tune in 4400Mhz CL19 anyway).

 

I would like to know which is better?

4000cl16 17 36
4200cl17 18 37
4300cl18 18 38
or this last test?

and to say that for xmp at 4400 it starts me up but it is not stable the reboots fail ... it will be a thing of the imc

 

going by the quick and dirty method, u get the following:

4000/16/17/36 = 0.408
4200/17/18/37 = 0.371
4300/18/18/38 = 0.349
3591/15/15/36 = 0.443

thus, 3591 would be best

or

4000/16 = 250.0
4200/17 = 247.1
4300/18 = 238.9
3591/15 = 239.4

with 4000 being best.

however, these are just rough calculations. in the end u have the check with your favorite software tools to see which clocks and timings give u the best performance. some software prefers high clocks / high bandwidth, others prefer low latencies / tight timings. others again dont really care at all about RAM

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

 

Thank you very much for your answers ... basically I use the computer to play ... how often would you use to play?

 

Nice! Is your chip delidded?

 

If you have copper delid and his, I plan to sand the die, to lower the temperature further

 

hello ,,, would someone advise ,, for some time I have a problem with accidental shutdown nb 7820hk ,, or in xtu test
I tried about everything, I can only think VR config"i use ac dc loadline" PSYS need setup ?? or do set up system agent ??? "The problem comes at higher voltages," reacts to temperature, but it's not the temperature

I completely rebuilt nb ,, is stable but ,, nevertheless higher setting turns it off at temperatures around 75C max below 105w config
i need about 10 points to first place on hwbot

,, but it is rather a possibility to set it fully when the temperatures are so good now ,,to improve stability, the confidence that it will not shut down
bios full unlocked
ec only stock
psu 230w AC DC set to 1 ..original get bad voltage spikes under load

----------------------------------------------
the original problem

lately I have a problem with "high temperature" but it's not about cpu or gpu or ram ,, nor ssd ,, so i don't know

,, vrm or uncore temp,.,. something ?? I don't know is simply not seen anywhere

,, the computer shuts itself down quickly. but never when playing or 100x load "stress test" rather on after a long time on the desktop and kicking into the load

certainly the temperature, with external cooling everything works, just can not locate where it is
all new pads ,, lq metal on cpu ,, own air tunnel ,, fan curve for testing I use cooler boost ,, temperature max 75 cpu x gpu 65, .with external blowing works fine even 5ghz, but now I can not get back 2 years functional 4.5ghz just do not know, after warming up a while before it can even be used without oc

does anyone know what it can be, I mean uncore ?? I used a single ram before and everything was ok
exist unlock for uncore more setup ??

spec my gt 75 7re psu230w
pl 100w cpu
42 x uncore
45 x cpu
refc-original 99.7
450 vr limit
2800mhz mem x2 original msi 16gb

----------------------------------------------
setup bios what i think is the problem
SA icc and voltage offset still shutdown
pll offset tested same

RSR ?? safe to disable ?
psys need reconfig ? i no
some setting of temperature or power limit

I really don't know anymore ,, bios ma perhaps all items ,, there is a lot here



I'm sorry it's a translator


thank any help .d


my gt75 7re



1589 is first ., but use 4.7ghz temp around 90C

 

It's your EC. An unlocked BIOS only does so much, as the EC is where the cancer resides. Unfortunately, I don't know of anyone that does custom EC's for MSI. Prema used to, but that was on the MSI 16L13.

 

You should talk with @Falkentyne

 

thanks,.

it must be the temperature when I use extreme cooling ,, everything is ok but extreme most temp get 20C cpu 40C under full load
I can change my settings in ___Thermal configuration___but I think I've already tried

,.it must be a temperature, if the machine is too hot too can not use my low profile
unfortunately the temperatures what are seen are ok ,, xtu aida hwinfo says nothing
,,ram an other now cooling custom fan .,more heatpipe on cpu .,complet air tunnel ,,all is cooling
yet at a higher room temperature 28C and above + testing or long indle ,.absolute brake nb ,,lower temp 22-24C work perfect

 

Just been messing about with my GTX 1070, seeing if reducing the Power Limit means I can run it at a higher core overclock - the theory being that perhaps high power usage might destabalise an overclock, and also based on the theory that at lower voltages you can run a slightly higher offset overclock (which can be seen by the shape of the overclock curve when you let NVidia do it's auto tuning overclock - it applies a greater offset to the lower voltages. (Actually, that could be an idea, maybe I'll run it through the autotuner at my reduced power limit to see what results it comes up with.)

Yes, so it looks like I can run 2 notches higher on core offset overclock Mhz frequency if I limit my card to 55% Power Limit. That power limit sounds very low, but I'm using the Gigabyte vBIOS on my Zotac card, and for some reason it under reports wattage by about 40W using the Gigabyte vBIOS, so it's higher than 55% Power Limit in actuality. I lose 2% in 3DMark and Timespy score, and in those benchmarks it's running just below 1V and just above 2000Mhz (bouncing around off the power limit). I tested it in BF1 where I run a frame limiter, and the frequency is actually higher in that game than it normally is - at the increased overclock and reduced power level it's anything from 2012Mhz - 2075Mhz (mostly towards the upper end and bouncing around in voltage from 1.00V-1.05V rather than being stable at a constant 1.05V), rather than a stable 2037Mhz. So performance is pretty much the same or slightly better, but using less power and at lower temperatures. I've not fully tested if for stability yet, and haven't tested it enough to know if it's gonna increase stutter in actual gaming, but on empty servers in BF1 I didn't notice any stutter. It's a curious little experiment, but it might be a keeper if I get the same performance for less power consumption/heat/noise. Not that any of those variables are an issue at the moment, but why not have them lower if performance is not really sacrificed.

(In the past I've tested going in the opposite direction too - increased voltage, and my card does not respond to anything above stock voltage - no increase in overclock stability, so that's not functional, and probably just goes to prove that it's at it's max usable voltage really).

 

Deep thoughts from Pcworld's PC expert...


Ask a PC expert: Should you buy an unlocked K-series Intel CPU, or save money on a non-K chip?

We answer a common question from our Full Nerd community!

PCWorld CPU guru Gordon Mah Ung has the answer for you though. He explains everything you need to know about unlocked K-series processors: Yes, they’re the only Intel CPUs that can be overclocked, but by purchasing one of these premium parts, you’re also committing to buying an enthusiast-class Z-series motherboard, as only those parts can crank clocks beyond stock speeds. K-series chips also lack a cooler in the box.

For those reasons and others, Gordon thinks most people shouldn’t pay extra for an Intel K-series CPU. That’s the short answer.

 

I think I fixed it .,it is the temperature below the mainboard ,.it does not seem to be a problem to complete the test with 30c rom temp it never succeeded ,.
moreover, if the lower temperature, added 0.05v ,, I do not know why but it is so ,.i get 20c down from pch ,,ssd ,.I'll wait for a colder day and try to break the record

just wondering what is still safe 1.35v? -- 4.7-4.8ghz

 

Msi Gt75 its an BGAbook. The higher voltage range will also affect the power delivery as VRM's etc, not only the processor. The outcome will probably be more like this beauty... Hardware Heaven!

 

Amps limit of 70 watts. No idea if that's EC or Bios but this is the Intel design limit for this processor. Even desktop Kaby Lake processors have this limit. Bypassing this limit will shut off the system.
VR Current Limit setting is not the same thing. Bypassing VR Current Limit will show Throttle: EDP OTHER or VR CURRENT in Throttlestop 8.70.
Desktop boards allow this to be overridden by the option "Current Protection". These laptops don't have this. Prema bypasses this protection in the prema bios but this is only for LGA boards from partner shops.

 

Higher temps will draw more Amps.

Try Prime95 version 29.8 build 6
https://mersenneforum.org/showthread.php?t=24094

And try 4.7 ghz on your processor and do small FFT AVX enabled preset and watch what happens to your laptop.

 

Hmm, that new setting wasn't 100% stable, it failed in Graphics Test 2 of Firestrike. I think the problem is that if the power limit does indeed increase the stability of the increased overclock, but the GPU only has to see one circumstance of load/frequency/voltage combination that throws an error - the power limit kind of puts a band aid on that, but it would mean lowering the power limit even more if I want this to be fully stable at the increased core overclock, it's possible that this will therefore become "not worth it" due to the performance loss. For example, 60% power limit fails immediately, 55% power limit fails eventually - I'd have to lower it some more. Might play with it some more another day - I bet 53% would be the magic number, just based on intuition, just might not be worth it.

 

hi Falkentyne

so my main problem is the "limit"., isn't it temperature?
in that case there is no point in trying to go above, my options end up somewhere around 1.3v - 1.295v, depending on the ambient temperature, 4.5ghz is also decent to play games, this setting is very close to the limit when only a higher ambient temperature is sufficient

I will try to go down or align the existing settings to make it usable on hot days

thanks a lot

 

Yep, 53% Power Limit was the magic number for what seemed like full stability at the higher overclock, however I tested it out in BF1 gameplay - it was not as smooth and felt like there was more stutter and mouse movement didn't feel as good. I think bouncing off the power limit whilst gaming is not the best idea, I think this is the cause. So, even though it allowed a greater overclock and less power consumption (about 40Watts less) for only slightly less fps, a reduction in the Power Limit was not worth it. That's the end of that little experiment, back to maximum Power Limit of 114% and my previous long time overclock (37Mhz less than what I tested here).

 

i9 9900k 5,1ghz uncore 47

 

How do you see this result?

 

You need provide more info, sir. What clock speed on Cpu and graphics. Here... Compare your results with equal hardware https://www.3dmark.com/search

--------------------------------------------------------------------------------

https://hwbot.org/submission/423144...0_(notebook_mxm)_21242_marks?recalculate=true


https://hwbot.org/submission/423144...80_(notebook_mxm)_7472_marks?recalculate=true

 

I think for 5ghz it's not bad ...

 

I planned to leave it that way, in games I don't think I lose performance from 5ghz to 4.9ghz

 

even at perfect scaling and CPU bound ud only lose 2%, so yeah...

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

 

After several tests I keep this: 4,9ghz 1,180v llc7 ram 3600cl15 .. most importantly without cache errors, 26 ambient degrees.

 

New benchmark in Gears 5

 

Geez, better than W7

 

I let my ZBook 15 stretch its legs a bit in my hotel room this afternoon. I see I need to update my AIDA64 Sensor panel and change the name to Geezer. Wheezer will be the MECH-15 since it is a BGA turdbook. The ZBook 15 is going to be a gift to my oldest son. His antique Dell Inspiron with a 2-Core Celeron and 2GB or RAM is slower than a 10-year itch, so it will be a nice upgrade for him.

https://hwbot.org/submission/4234212_

 

Anyone know how to submit a Geekbench 5 score to HWBOT? It's asking for a file. What the heck is a data file?

 

I don't have Geekbench 5, but you should find something like a button in the Geekbench 5 user interface to export the results after the benchmark is finished. Save that file to your desktop and then upload it at HWBOT. If you do not see a clearly marked button, if there is a top menu, look there for an option to do that.

 

Many thanks to everyone, I wanted to comment that with this voltage of 1,205v 5ghz when I play BF5 the computer restarts after 15 minutes, although realbench passes 1 hour, to solve the problem I had to raise the voltage of the cpu to 1,215 I've been playing 4 hours and everything is fine without problems, what I mean is how I spend realbench 1 hour and instead the computer peta me when I play BF5? Does it have explanation?

 

BF1 is quite tough on the CPU, so I guess BF V is too. When testing for overclock stability it's probably better to use more than just one source. I use different stress testing programs depending on how 'stable' I want my overclock to be. I always test with Prime 95 v26.6 which is my test for non-AVX stability. And depending on how stable I want my overclock to be then I'll choose between x264 on a loop (from here: https://www.overclock.net/forum/5-intel-cpus/1570313-skylake-overclocking-guide-statistics.html) for a relatively easy pass, or OCCT to test for absolute stability. OCCT is the quickest & harshest test to find errors, and it's the criterion I use currently to validate my overclocks - it seems to offer more 'predictable' behaviour in BF1 with less "weird stuff" happening in one on one engagements. The amount of voltage difference required for my CPU between an overnight x264 pass and a 1hr OCCT is anything from 50-75mv if my memory serves me correctly - obviously more voltage required for OCCT.

 

This is good advice.

Another couple of points that I have found with games. After years of being an overclocker first, gamer second, I have found that settings I have used for many months for benching without a single hiccup sometimes act totally spastic with a game. BF and COD and Crysis games are among them. All of those titles are more CPU heavy than normal as well. Bumping BCLK above 100.00MHz (even as little as 100.25-100.50) and getting too aggressive with ring/cache ratio overclocking can manifest in weird problems that may otherwise never surface when you are playing certain games. Same can be true with memory overclocking. I think some games just get buggy when things happen much faster than anticipated under normal scenarios, and when things happen too fast it presents an opportunity for error to occur. Some software handle errors more gracefully than other software.

Chrome browser is another non-game example of software that it not very fond of running an overly aggressive memory overclock, especially if browser hardware acceleration is enabled. I have had many times when I have tested settings for several hours and had no errors or crashes only to find I couldn't launch Chrome without problems if hardware acceleration was turned on in the browser settings.

 

New XTU version. But I don't think there will be a change...

https://community.hwbot.org/topic/189976-xtu-will-become-pointless/?do=findComment&comment=539754

 

very good occt and realbench .... all good with these programs and I leave it stable, but I need other programs similar to when you are playing 3 or 4 hours ... with occt and realbench I leave it stable but after 3 or 4 crash hours ....


I put voltage 1,215 and in 4 hours crashh

 

Are you saying that you can run OCCT for 3 or 4 hours and it's stable, but then it crashes when you game in BFV for 3 or 4 hours? I'd be surprised if that was the case because OCCT is way more strenuous on the CPU than BFV, but either way if it's crashing in BFV then you either need to increase the voltage or reduce the Mhz. Or perhaps your RAM is unstable, and maybe OCCT doesn't stress the RAM as much as BFV - so you might need to reduce your RAM overclock. But, I know you've been testing your RAM overclocks with HCIMemtest (I validate with that), and in my experience that is super stressful on the RAM, so that makes me think your RAM is stable. You would need to experiment with more voltage or lower Mhz to dial in gaming CPU stability, and on the flip side you could experiment with a lower RAM overclock whilst leaving your CPU at your current overclock - that way you're finding out if it's the RAM or if it's the CPU causing instablity whilst gaming.

In my experience 1 hour of OCCT stability equates to total stability in all gaming scenarios, so I'm surprised to hear what you're saying (if that's what you're saying). In fact even x264 stability (lower voltage) resulted in total gaming stability.....but I did get the 'feeling/impression' that BF1 wasn't giving me kills in situations where I expected to get them or sometimes it 'felt' like enemy players were sometimes 'glitching' when I was shooting at them and therefore they were getting away more often....so I can't really call that instability with 100% proof...it's more of a feeling that the game runs more predictably when I use a higher CPU voltage.

 

sorry use translator ... xd I meant to spend 1h of realbench and 1 hour of octt ... to make it stable ... but I see that you are right bf5 pulls more cpu than these programs, which program can be better to leave a stable system? burn the cpu ... xd

I play with the ram at 4300 cl17 stable with hci

I do not know, if it is too much to play at that frequency of ram or should I lower it ..??? Because from 4000mhz to 4300mhz neither boy to lose fps? or performance?

 

Only testing in your workflow/games will give you the answers. But going from 4300 down to 4000MHz on ram shouldn't reduce FPS. Still instabllity, I would lower Cpu overclock first to see if this have an effect. Then do the opposite... Use your max stable Cpu overclock but reduce ram speed.

 

Ah right, so your overclock is indeed 1hr stable in OCCT but not stable in BF V over a number of hours - I'm surprised to hear that, but it is what it is. Well, in that case you still have to experiment to gain stability in BF V - like both me & Papusan already suggested try to find out if it's an unstable CPU overclock or an unstable RAM overclock.

Try two things:
1) Your current CPU overclock with a lower RAM overclock (use a very safe RAM overclock, maybe even 'stock XMP')
2) Try a reduced CPU overclock (lower the Mhz, keep voltage the same) with you current 'high' RAM overclock

Doing the above should show you if your BF V instability is caused by the CPU or RAM, and then once you know the answer to that you can start working on gaining stability for the correct variable.