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

Cool, what components have you gone with?

 

Wrong ones from what I see online

Anyway the fist stab at the build is ready

Intel 9900k under AIO Asetek 645LT
Asrock z390 Phantom Mini/ac
2x8 GB Corsair Vengence LPX DDR4 3000
Phanteks Evolv Shift
Silverstone SFF-L 800
Nvidia RTX Titan

Temps are under control, noise is decent

I need to chnge some things as I am not happy with the current cable management and the place where I was obligated to mount the AIO (as the tubes are short not bought for this case) also my goal is < 10L case ... this one is too big )

 

Nice main components! That orientation of the GPU looks unusual? And yeah, that's nearly a normal sized midi case (?), so I can see why you want a smaller case considering you got such a small motherboard?

 

What do you mean normal sized midi case ? we are in the SFF territory here

 

Yeah, I dunno, that's why I was asking you, to me it looked like the case was standard midi (didn't look very small!), but I wasn't sure, hence the ? mark. I also noticed the GPU is orientated unusually, not seen that before.

 

The case is ~ 22L in volume ... you can check here some of the "competition"

https://docs.google.com/spreadsheets/d/1dzRY3LLsXcDX5Qt5PvQOrAJhw0EK9Cy7UtQPdyRp6PY/edit#gid=0

The orientation is what is recommended by Phantex I know is weird (I can swap it out and reverse it, but I want to have a baseline from temps perspective first)

 

Oh well, this didn't work out in the end. A 3 thread load of OCCT would destabalise my 100Mhz extra overclock boost that I'd gained through Per Core overclocking. My conclusion is that Per Core overclocking is only useful if you're thermally limited rather than "silicon quality" limited - because you wouldn't have the thermal overhead to overclock all cores to their theoretical max stable value, so you'd compromise by using Per Core overclocking to reach the limits of the "silicon quality" with say 1 or 2 cores, and then you'd have to lower your All Core overclock to stay within thermal limits. As it stands though, I think Per Core overclocking is only useful if you're thermally limited, as it can't magically override the "silicon quality" aspect of overclocking.

TLDR: Per Core overclocking has given me an extra 100Mhz in single and two thread loads vs All Core overclocking, and a corresponding small boost in performance: it could be worth your while to experiment with Per Core overclocking vs the traditional All Core overclocking.

Ok, this might be interesting, what are your thoughts about "Per Core" overclocking in a desktop PC environment rather than using "All Core". Per Core overclocking you set different multiplier ratios depending on the number of CPU cores active in the load. So my latest All Core overclock is 4.69Ghz @1.4V (OCCT 1hr stable), but I figured I'd try "Per Core" overclocking to see if I could get one and two core loads to be a couple of multipliers higher whilst keeping 4 core loads at my previous 4.69Ghz. I'm testing it right now, I'm testing 1 thread of Prime95, and the cores are getting up to 4.9Ghz! It's not crashed or thrown errors yet, but I've only been testing it for 10 mins.

I'm thinking it might be harder to test for proper stability if you're doing Per Core overclocking? What are you thoughts on the value of Per Core overclocking, and your thoughts on how to properly check it for stability? (I'm familiar with how to check an All Core overclock for stability).

I've set 4.9Ghz for 1-2 Cores, 4.8Ghz for 3 cores, and 4.7Ghz (well 4.69Ghz) for 4 core load - this is all at my previous 1.4V.

I'm thinking that we're actually leaving performance on the table by only doing All Core overclocks, your thoughts?

*EDIT: When looking in HWInfo Sensors whilst I'm doing the 1 thread load stability testing I can see that the 1 thread load jumps from one CPU core to another - cycles through them. So average CPU load is balanced out throughout the cores to about 25% load each. That's on Prime95 non AVX, but on OCCT it seems to be intelligently focussing the load on the coldest CPU cores. Hmm, that means the OCCT load testing is ignoring the other CPU cores effectively, so they're not being stability checked - I think this makes the OCCT testing invalid for stability testing because it's ignoring the hotter cores. I'll try Prime95 AVX to see if it shows the same behaviour - I need the load to cycle through all the cores, otherwise it's not testing the stability of each physical CPU core. Either way, I think truly testing stability will take longer when Per Core overclocking because the load is only about 25% per core on average, so I think it's gonna take 4 times longer to thoroughly check for stability of an overclock - my thoughts anyway.

**EDIT#2: Well 4.9Ghz for 1-2 cores wasn't stable, Prime95 threw an error. I'm thinking that perhaps there isn't any more performance left on the table by just using an All Core overclock - it might just be it takes longer to find errors when stability checking Per Core overclocks. So, I've backed it down to 4.8Ghz for 1-3 cores, whilst keeping 4.7Ghz for 4 cores. I'm testing it again now, so even if it passes a stability check it's only an extra 100Mhz gained by using Per Core overclocking. LOL, there's probably a high chance I'm corrupting my OS with this testing!

***EDIT#3: I've tested stability of the Per Core overclock overnight with Prime95 AVX running on two threads and it's stable, with an average CPU clock of 4.76Ghz per core as measured in HWInfo Sensors. This is at 1, 2, & 3 Core at 48 multiplier, while 4 Core is at 47 multiplier. I also validated with CPUz, and single core performance has increased in line with the extra 100Mhz I was able to attain through Per Core overclocking as opposed to All Core overclocking, I've extended my lead at the top of the table! ;-) https://valid.x86.fr/t02hmr


TLDR: Per Core overclocking has given me an extra 100Mhz in single and two thread loads vs All Core overclocking, and a corresponding small boost in performance: it could be worth your while to experiment with Per Core overclocking vs the traditional All Core overclocking.

 

Just picked up this bad boy off Amazon "Renewed" for $1545 out the door. Just got it about 1 hour ago, testing things out and I must say I am grinning from ear to ear. This little device is amazing for the power it packs.

1080p 144hz 3ms with MUX -- Optimus and G-Sync available via the software and restart. God I hate Optimus so much, the difference just browsing the web and tooling around on the desktop is so much smoother and responsive.
8750H -- sure last gen but honestly tiny difference
RTX 2080 Max-Q
16gb in dual channel
512gb NVME
Super thin and light weight to carry around with me through airports

https://www.3dmark.com/3dm/40356996 -- 80w vBIOS with +800Mhz memory and +125mhz core. Will flash the 90w TDP later. Also need to flash the BIOS with unlocked to get access to memory so I can install my 3000Mhz CL16 stuff. Unbelievably this thing doesn't throttle the CPU down under load in Cinebench. It ran 3.9Ghz all day with the -150mV undervolt I applied to it. Temps with max fans were 70-80s. Will be doing a LM repaste to CPU and Gelid Extreme to the GPU eventually.

https://imgur.com/a/KI5dHlw

 

Alienware m15 R2 after undervolt and some airflow under the chassis:



"Out of the box" :

Spoiler

Honestly I am not sure what else to try. You can try the XtremeG drivers as well, there's a thread here in the graphics card section.

 

https://imgur.com/a/6Xe0s3Y

Predator with liquid metal on the CPU sitting flat on my kitchen island.

 

New top scores with the Tongfang killing everything else with the same specs... all #1 for 9750H and RTX 2060...

 

Mr. Fox is back again on (soldered) Notebooks!? Did I miss something?

Gesendet von meinem CLT-L29 mit Tapatalk

 

You might have. I got this a little over a month ago for work/travel because it was cheap, will play games OK and is the best value in a turdbook. I don't like that it's a turdbook and it has all of the features that I hate about turdbooks, but it offers the best bang for the buck for less than half the cost of what I would want in a laptop. I am not willing to spend $3,000-4000 or more on laptops any more. Those days are over. They're just not worth the hassle when a desktop is a better option, and they take way too much effort to make them run right because the people that make them do everything half-assed just like they do the turdbooks, and they intentionally do things to make upgrades an inordinately costly and inconvenient pain in the butt.

I'm not benching the desktop any more only because I've maxed it out and cannot achieve more than I already have with this hardware. The last few times I tried, I totally wasted the better part of a day hitting numbers that matched or fell just a few points short of what I had already done. Until something increases in terms of specs or thermal configuration, there is no point in pursuing higher scores because I'm not getting any higher scores than I already have. I still love the desktop though. It totally shreds everything that gets in its way.

 

Are you home now? Hope everything is well with you Brother. Sent you an EM

 

Hey bro. Happy Saturday. Yeah, was home this week. Leaving tomorrow on another business trip. Got to replace the alternator on the wife's Sorento today so she will have her wheels while I am gone. I absolutely despise working on cars, but I hate paying someone else to do it just as much. How funny is it that an ASE Certified Master Auto Tech hates working on cars, LOL. But, I do... it sucks.

Edit: hmm... check your drafts folder. I don't have an email from you, bro. I'll recheck after I am done with the alternator.

 

A performance comparison between Redmond's new and shiny Win 10 vs old Win 7 on the one core Celeron M743 from the old days. New is as always better

Windows 7 (965-909)


Windows 10 (941-890)

 

Newer is always......... newer.

 

Prime95 version 29.8 Smallest FFTS 1h , 4,9ghz.

 

Sorry to ask question here again brothers..Please feel free to give any of your inputs <anything> as It's invaluable.

As I mentioned way back ( Feb 2019) about my interest in buying a PC this year.. FF now, with PCIe gen5 / DDR5 coming in 2021 and BIOS EOL in 2020 plus Win7 EOL in Jan 2020, and Intel 14nm woes along with AMD's dependency on the mobile optimized TSMC 7nm node. I'm very torn on the purchase of a new beast.

I checked OCN forum (AMD motherboard section) just a few pages skim and watched all GN videos except the TDP one, covered DF ( To recap - 3900X vs 9900K & 3700X vs 9700K), Der8aur as well. First of all that ABBA BIOS whatever it is is totally unstable on Gigabyte boards (OCN link), it ruins the DRAM OC potential and it's not going to guarantee that Ngreedia style GPU boost on the AMD Zen2 with a ton of other shenanigans like It needs Ryzen Master 100% and the marketing confusion BS. To make it worse there is still no damn datasheet WTF. Plus there are higher voltages all over the place, even the Trident Z Neo has higher DRAM voltage (Not that it's a major concern) all new GSkill RAM kits optimized for AMD using B-Die are of higher voltage than the earlier batch. Next up the fclk, mclk, uclk all those settings and DRAM optimization is heavily dependent on BIOS and IMC as usual and Mobo too so XMP might still not work for the high speed/tight latency kits. Add the High PCH temps on the X570 chipset and RPM for that fan is very high, Gen4 SSDs absolutely insane temps on a desktop board.

Ultimately I could be wrong on some points, since I just skimmed a few pages, the X570 is still in WIP (? Reviews are out since July 2019 & worse if you pair with the retarded pile of dump Win10) because I even wanted to buy the ridiculously expensive Aorus Xtreme at $700 (OCN link) (because I think there won't be an X670 chipset again since these are alr full of features like 10G LAN and all other bells and whistles, the Gen4 etc plus no chipset fan and the ESS9218 DAC chip same as in my LG V30) and still end up with issues and non quality BIOS, I know GN used Aorus Master for their charts..vs a Z390 platform rock stable, DRAM 4000MHz no issues and can even push more on tighter timings. Plus the Z390 EVGA Dark is very tempting. Along with the binned 9900KS (If I can get a hold of it), bonus is Z390 Dark takes Win7 too vs the LTSC 1809 while AMD needs 1903, I think I will choke if I use that 18Month EOL and 6 Month update cycle. Gen 4 SSDs not much useful I guess, and saturating an X8 link itself takes a lot of datarate perhaps even 4K for 2080Ti, X16 3.0 needs much more exponential gains, however it is future proof to an extent.


I even bought the DRAM kit (4x8GB - 32GB) F4-3600C15D-16GTZ (Fastest and Tighter B-Die apart from F4-4000C17D-16GTRS, ) when the B-Die EOL news came plus even the kit which I got was always going out of stock on Newegg and Amazon doesn't have it at all (Bought around Aug-Sep). And we still have G.Skill new DRAM kits, and also to insult the C15 Trident Z which I bought is readily available as well.

Choice of CPU - 3700X as Ryzen 4000 final AM4 CPU would be the replacement for the 3700X plus 3800X, 3900X perf in games is very less variance of ~6%.

Case - Lian Li O11 Dyn XL or Corsair 500D Obsidian or the overkill Phanteks Enthoo Luxe 2 (tons of Storage bays ! I already have 4x10TB WD REDs with me will get a pair of Samsung NVMe 2TB 970 Evo Plus + SATA 860 Pro)
Cooling - EVGA CLC 360 or Noctua DH15 (ofc won't work with Z390 Dark),
Monitor - ASUS ROG PG258Q or Acer Predator XB272 (FHD, 240Hz, GSync, 24") with a couple of Noctua Industrial fans
PSU -Seasonic 1000W Prime Ultra Ti.
GPU - ASUS ROG Strix 2080Ti. I need all the performance and a cool desktop with minimal bling.


To sum it up

- Unstable BIOS always needs fixing.
- Must use Win 10 1903 (WTF because no LTSC won't save that Zen uarch on Win10, and even the future LTSC no idea when It will come and might need more fixes always needs updates..)
- Zen 4000 drop in upgrade (Perhaps will beat 9900K finally ?, because I really care about gaming since I have a ton of old games which are great vs latest PC drivel nonsense) will the BIOS still be unstable bonus pack Win10 ?

vs

- Stable platform Z390
- Z390 is EOL, Biggest blow to be honest, because if this was not the case, I would have already ordered all parts. Comet Lake 2020 Q1-3 LGA new pin 12xx, Wonder if 9900K/S is also EOLed by that since pin count is different for the final ring bus SKL uarch CPU. So sad..

Or

- Ugprade to a 1070MXM and 120Hz (Already there) and be done with it until 2020.

But with all that drama going on I'm really lost as I don't have any idea now..
Is AMD platform still unstable as I read ? don't want to spend a lot of time this holiday season, esp the work was eating me out I barely check NBR now. Shame.
Will still dig through for more info.. Thank you !

 

A lot of people are using Ryzen 3000 CPUs and from what I can tell most are extremely happy with them. The Ryzen 4000 is expected to still work on AM4 motherboards, so there would likely be another upgrade path.

 

I myself will wait to see Intels new gen processors before I make a move to new computer. If the old is still good enough I can't see a reason jump on old. Neither is todays AMD tempting. Maxed out clocks or small oc headroom ain't fun either.

 

I think it's a bad time to build a desktop now - NVidia 3000 series coming out next year on 7nm / Intel supposed to have something new too (not sure when). If you build now you're buying everything that's right on the end of life as you say. If you've got something that can game ok at the moment, just keep that until next year? You talk about upgrading to 1070MXM, so sounds like you don't have adequate hardware at the moment, they're not cheap though I'm guessing and is that Ok to spend that for just a years use? I'm certainly not building anything this year, but I've got something to game on and use which is completely adequate for me at the moment. If CPU gaming power for high refresh rate gaming isn't your highest priority you could just buy one of the 3000 series AMD CPUs which has lots of cores for any future games (as they become more multi thread aware) as well as good for productivity tasks, buy a cheap second hand GPU from ebay to tide you over till next years NVidia release.

 

Spoiler

20C core temp difference at max delta
That chip needs a delid more than I need new underwear....

 

@Prema Mod Team Today

 

What Loadline Calibration are you using?

You should not be pulling 167 amps at 1.243v load Vcore (Maximus XI measures vcore with differential on-die sense).
I get 143 amps at 1.264v (at 5.1 ghz), even though I only use the defaults (4 gb RAM) for Realbench, shouldn't make a big difference. Using 1.375v Bios set with LLC: High (0.8 mOhms loadline) for a healthy vdroop (Vdroop is good for your processor).

I'm asking because i notice your CPU Vcore is literally 12mv vdroop between idle and 163 amps load.
This looks like you're using Level 8 Loadline Calibration (0 mOhms)

You're not using LLC8, are you?

 

if I am using llc 8 ...
What do you see wrong?

 

Looks good to me. I don't see any issues. Having less vdroop is good. I have vdroop disabled on the X299 Dark and with the Maximus X Hero I ran LLC at 10.

 

@Mr. Fox I did have the email I sent you in my drafts folder. Anyway I sent it to you . Thanks

 

https://hwbot.org/submission/4273692_papusan_cinebench___r20_core_i7_920xm_1054_marks

 

Anyone getting the 9900KS? It's sold out at the moment, but I'm curious just how 'binned' it is to see its potential past 50x.

It's said to run 50x at pretty low voltages.. Hmm..

 

https://www.3dmark.com/fs/20861606 | https://hwbot.org/submission/4274017_

 

1740Mhz on the GPU there during the runs? You're for sure limited by power and maybe temperature rather than the silicon then, is it worth doing something to up the allowable GPU power or are the temperatures not good enough and the GPU power supply not sturdy enough?

 

Ah you just reminded me of my M15x, only pics I have left of it...



https://www.3dmark.com/3dm11/5482014

 

Yes, 2060 has severe power limitations. Power limit is set absurdly low (80W) and should be set to at least double that. Voltage is too low as well, maybe to a slightly lesser extent. From what I can see this is almost entirely induced by cancerous firmware. From what I can see, it's all 2060 (desktop and notebook). Probably so 2070 results looks stronger in comparison. Typical Green Goblin smoke and mirrors nonsense.

Running with the laptop on AC cooling doesn't remedy the low core clocks that are basically "power limit throttling" similar to when a CPU has the same problem. Under benchmarking loads the total system power draw is well below the AC adapter's capacity.

 

You can't use a hardware flasher to flash the GPU with a modified vBIOS? I think I heard that's not possible. You couldn't do a shunt mod on the GPU somehow so that it under reports GPU power?

 

GTX 460M can't compete. But the Cpu in the old Clevo lady can https://www.3dmark.com/compare/3dm11/5482014/3dm11/13552683#

 

I haven't figured out which resistors I can shunt. That would likely almost eliminate the need for a firmware mod. (Firmware mod would still be useful for higher voltage.) If I can figure out which resistor(s) to shunt, I certainly will. They might be on the keyboard side of the motherboard. I haven't been able to spot them on the bottom side yet.

I did try applying modded firmware with my SPI flasher and it programs fine, but puts the GPU in a code 43 state and it won't function. I might have set the power limits too high or modded the firmware incorrectly, but I got tired of screwing with it. The vBIOS chip is under the heat sink, so that makes it a pain in the butt to have to fart around with. It does a fine job of playing games and that's really all I expect from it other than using it for work. It does a fine job of that as well. If/when @Prema has time to tweak the vBIOS I will try that to see how it works. Otherwise, I will just keep using it for the purpose for which it was purchased.

I had a good time playing Gears 5 with NVIDIA Surround enabled last night. Plays surprisingly decent at 3840*1080 with ultra presets... 2060 delivers between 60-75 FPS consistently, even with that high of a resolution and ultra presets. (At 1080p it is generally 90-110 FPS with the same settings.) I really like my new AOC C24G1 144Hz VA panels. Finally retired the old ASUS 60Hz panels with bezels too wide to game with Surround enabled.

 

VA Panels are really good for excellent contrast ratios and deep blacks and good shadow detail, whilst being a bit smeary/slow for competetive high refresh rate gaming, so I'm sure it's a stunning visual experience for general gaming, and surround displays add to that - the central bezel might annoy me so if I did that I'd probably have 3 monitors. I kind of yearn for the visual gorgeousness of a quality VA panel, but my need for speed requires me to have TN, but I do yearn!

That's impressive performance from that 2060 regardless of the power caps. Yeah, well if you feel like trying to do shunt mods & vBIOS mods on it then I can see the worth of that, but only if it remains practical, you don't want to turn your functional laptop (the reason you bought it) into a frankenstein machine that requires it to be operated outside in Winter or over an air conditioning unit! I suppose if it was duel purpose in as much as you could limit how far you press the throttle for everyday stuff and then have an easy accessible benchmarking mode, then that's still perfectly practical. Last thing you want to do is turn your laptop into a "pain in the ass"!

 

Some info
https://wccftech.com/intel-core-i9-9900ks-silicon-lottery-pre-binned-cpus-available/

If you have chips from older batch with P0 stepping and good bin, you won’t miss much.

 

Indeed. I'm just curious to see how well it OC's with lower voltages, however, like I've discovered on the 8086K for example, although it too is a binned chip, I've found some to require slightly more voltage at higher frequencies than some binned 8700K's that oc'ed better @ 53x and 54x, while it required higher voltages at lower frequencies below 50x.

ie... I wonder if the 9900KS chips are binned only to thrive at 50x and am curious how it'll handle higher OC's.

 

The R0 stepping to mitigate some of Intel’s security flaws will eat of IPC. Of course higher default clocks was needed In same way as 9750H(R0) vs previous gen 8750H BGA chips (don’t offer much with slightly higher clock speed).

If you remember. The last special binned chips (8086K) from Intel offered same 6 core boost as 8700K. Imagine if newer KS chips would run same all core boost as it’s brother

 

Yeah, I agree. Having the BIOS unlocked is awesome and I would like to eliminate the power limits just so the GPU can function at its full performance potential like the CPU is doing now. The access to memory settings and other tweaks (like BDPROCHOT, SPD write access, power management, etc.) is also really nice. The temps are OK and Tongfang did a better than average job of designing the cooling system. But, still... it could be better if it were a little bit thicker and heavier. Pushing it harder with an unlocked CPU and GPU running at full potential would make it less functional using its current thermal management components.

As far as these panels go, their specs are great. They have a 1ms response time. Cannot beat them for the price... very reasonable. I am also able to set them to 2560*1440 @ 144Hz with less decay of the visual quality than any other monitors I have seen before. The quality is nearly as good as the native resolution of some of the crappier monitors that I have used before. Of course, that isn't playable with a 2060 and high quality settings running in surround (5120*1440) so there is no reason to even bother with it. But, it says something about the quality of the panels that they can actually do it without going to blank black screen or looking so horrible that it makes you go cross-eyed. And, if a person actually cares about G-Sync it works on this FreeSync panel.

AOC C24G1 24" Curved Frameless Gaming Monitor, FHD 1080p, 1500R VA Panel, 1ms 144Hz, FreeSync, Height Adjustable, VESA, 3-Year Zero Dead Pixels

https://us.aoc.com/en/gaming-monitors/c24g1/specs

In a third-person shooter title like Gears, the center margin (no bezels are present) is narrow enough that after about an hour of gameplay I more or less got used to it and essentially forgot it was there. In a first-person shooter title it would be more distracting because the crosshair is perfectly centered and split between the two displays. With RPG, Strategy and Racing titles (which I am not particularly fond of) it would be even less distracting.

 

Very little to complain about with my 9900KS.

 

Looks like the photos didn't make it along for the ride.

Ah, much better. Looks like you got a really nice sample. Less than 1.300V at that clock speed is pretty sweet.

 

Bro Fox. The eGPU spam is even in the 3DM leaderboard for 10 years old graphics as GTX460M

 

Yeah, that really sucks. It makes it difficult to sort fact from fiction when 3DMark cannot even properly identify the GPU handling the workload. I've seen that numerous times and it is very misleading.

 

And nothing will be done. At least the bot doesn't have same mess.

Not bad from a laptop from 2015. New ain't always better

https://hwbot.org/submission/4274344_papusan_cinebench___r15_core_i9_9900k_2241_cb