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

I get the code 92 if I try to boot my 3200 RAM above 3500. I figured it was just the RAM was too weak to handle over 3500, but maybe it's not the RAM. It runs 3400, 3466 easy and eventually boots 3500 after a couple of tries and having to press F1 after failed POST.

Edit: I should try a higher RAM overclock with this newer/better CPU. It might work with this one.

 

Yeap good quality 6 ft DP cables that i have been using for ages with 0 issues.

DGPU. When connected to iGPU, they work fine. Or if the iGPU is completely disabled on Z370.

Not watching the temps was probably my biggest mistake, even after reading about so many issues with over heating VRM's. I had it at LLC 6.

That one reason why i straight away went with the EVGA x299 Dark . Now just waiting on the 7980XE, and the delid kit.

Unless the x299 Rampage has active/proper fins with heatpipes cooling on the VRM's i would highly be cautious with heavy over clocks.

Weird, my 64Gb 3000MHz CL14 Trident Z kit can boot at 3800Mhz CL17 @ 1.35V . But not really stable, since the nitty gritty tweaking hasnt been done. I havent had the Code 92 on the Code even once so far.

 

Yeah, it might have been that previous crappy 8700K sample that I took out. I need to try it on this new CPU and see if it fares any better. I think I will do that here in a bit and see if it works.

These came today, so maybe this weekend I will work up the nerve to do the 1080 Ti mod.

 

That was all they had in stock so I grabbed it while I wait on my real board to show up. I have a OC Panel that I use so I can only buy certain boards to run it. I liked the Apex better over the Extreme, so I grabbed that one.
Good call on the advice to @Trafficante

What does your MB book say code 92 is for you?
And It could be the ram as well. At this point I don't know. I'll know tomorrow and then make a judgement call. Considering it was doing it without the ram even being over clocked makes me wonder, BUT!.....yes, there is a but. LOL.
I did not have one single error what so ever with ram on the Apex board. And it booted my XMP ram first try. And stayed running from that point on.

 

Sweet. I would also read up on Tin's uncorking writeupt that i linked to in that earlier email. Definitely super useful.

 

Same as what you saw for your board... "PCI Bus initialization is started"

 

I'll get to the 7980XE a little later I think. I need 7940X/7960X points first. And I might have to find a 7900X as well.

 

No go on the RAM still. But, now it is not code 92. It is code 2b - "Memory Initialization" if I try to exceed 3500.

 

Which memory kit do you have exactly?

 

@Mr. Fox

Oh, sounds like Team Premamod benching session at Brother Fox's house this weekend.
Got my gear and ready to roll!





Not with that board of course. LOL
@bloodhawk
Gskills maybe?

 

Sounds like fun. Been a long time. What time you going to be here?

What is this HWBOT achievement? S441?
JOHNKSSS
Achievement gained: Get 30 Hardware Points on S441

 

It's the cheap "gamer-boy" stuff, which is all I could afford. 3466 is the most I can get it to go stable.

 

Possibly 6 am Saturday Morning.
Socket 441 Motherboard.
http://hwbot.org/achievement/3081_

 

So, you're leaving about midnight tomorrow?

Actually works better at 3400 using my own XMP2 profile with tighter timings than 3466.

 

Yep!
You going to be free the whole weekend?

 

You bet. Would love to have you here.

 

Great! I'm bringing everything.

 

Ah, but even these cheaper kits sometimes clock better the more expensive b-die lower latency chips. There are a lot of misconceptions around regarding better overclocking memory. End of the day its just better to grab whatever highest clocking you can and push it from there.

Be sure to take pictures!!
Would love to see what crazy setups you two will be churning them points on!

 

You saw mine. LOL

I need to find me another bench station though

 

Make sure you are using dual ranked kits at least. Otherwise DDR interleaving would not work and your actual memory speed drops. Memory speeds is more than just clockspeed and timing.

Also nice to see you and @Johnksss have some fun overclocking. High overclocking may not be for me but definitely nice to see.

 

That HardOCP review where they didn't see much performance difference, I think the guy said he was gonna do some more revisions to it, can't remember the revisions, I think he talks about it in the video on that Guru3d link I posted. About the 20x closer thing, I'm just quoting what he says in the video. To me it seems like a good idea & design, and that's why I linked it up on here, but we'll have to see how the finished product performs to really guage effectiveness - hopefully he'll reach his funding target.

 

wPrime 1024M @ 5.4GHz - http://hwbot.org/submission/3826964_

Cinebench R11.5 @ 5.4GHz - http://hwbot.org/submission/3826962_

Cinebench R15 @ 5.4GHz - http://hwbot.org/submission/3826961_

 

https://www.3dmark.com/spy/3569460 | http://hwbot.org/submission/3827243_

 

Absolutely. Would be awesome... come on over.

 

My tubing and fittings finally arrived earlier this week, finally ready to start finishing my build, super excited to get this finished and start testing and start over clocking with you guys. Also I met this awesome glass blower twitch streamer who is local to me and I am going to work with him once I have my bends and tubing done, to try and give him my tubing to make me some cool custom glass tubing for my rig. If all goes well with clear glass and things work out, then I plan on re doing it with some cool custom worked glass with wigwags and UV glass etc. I think it will be super awesome if we can get it to work out but my priority is to get it up and running with PETG tubing first, then give him all my tubes so he can try and re-create it with glass and know the correct bends and sizes he needs to do. Shout out to https://www.twitch.tv/pokeranger69 if any of you are interested in that kind of stuff, he does a very good stream and very interactive with chat explaining what he is doing etc etc. https://www.instagram.com/pokerangerglass https://www.instagram.com/chrisinchaosglass Hopefully the glass idea works out, I think that would be an awesome/unique addition to a build that not a lot of other people can say they have done.

 

Sounds great, will be good to see how your build comes on, plus the custom glass tubing sounds really unique too.

 

Yeah it will be a bit down the line that that eventually happens but I am super excited to try to work with glass. Originally pre cut tubes/pre bent tubes of glass were quite expensive and not very long and would have required a lot more planning and my buddy stopped streaming for a little during the time I started this and I lost contact with him. Hoping now this idea can come to fruition, even just clear glass will be cool but if all goes well, I cannot wait to have some sick glass in my PC, especially if we can do some cool stuff with UV reactive colors so it illuminates under light, I have so many ideas hopefully I don't end up getting dissapointed haha.

 

Results from AMD's new chips ticking in. Awful reading... http://forum.notebookreview.com/thr...ga-polaris-gpus.799348/page-424#post-10707641

 

I've benched my ram at 3600mhz with Ryzen, and it's really picky about ram! To be fair it is 3200mhz stock

Sent from my SM-G960U using Tapatalk

 

I'm putting absolutely zero faith in those leaked benchmarks, as I think they are fake or not legit. I will wait for April 19th.

Sent from my SM-G960U using Tapatalk

 

I am pretty sure it's this crappy Hynix M-die RAM. I've seen Maximus X Hero boards rocking 4000 RAM already, but they're not using the el cheapo G.SKILL gamer-boy sticks.

 

If I remember correct, it was run livestream as well.

It's what it is.
[UPDATE]AMD Ryzen 7 2700X Overclocked Up To 4.3 GHz (1.4V) Across All Cores in First Preview Testings

[Update]AMD Ryzen 5 2600 CPU Overclocked Benchmarks Posted By User - Max OC Potential 4.0 GHz, Faster Memory

 

I still have no faith in someone that violates an embargo. The numbers for wprime aren't any better than mine. In fact I can beat the 32m multi core score at only 3.7ghz?? Similar score on single core also. If it boosts to 4.35, then the scores should reflect better than what is shown. The CB score should be higher also. Guys in my league are matching that score with 1700X's, and 1700's with less clock speed.

http://hwbot.org/submission/3731750_abilioalbino_cinebench___r15_ryzen_7_1700_1963_cb/

Sent from my SM-G960U using Tapatalk

 

It' will always come leaks. And some are trustworthy. And we can also do some math.

Compare stock clocks (Cinebench R15). And Wprime is a difficult task for AMD (new chips). Compare results with Intel and you will see

 

Time Spy: https://www.3dmark.com/spy/3577385 | http://hwbot.org/submission/3828453_

Fire Strike: https://www.3dmark.com/fs/15279036 | http://hwbot.org/submission/3828568_

 

Fire Strike: https://www.3dmark.com/fs/15284270 | http://hwbot.org/submission/3829099_

Fire Strike: https://www.3dmark.com/fs/15284348 | http://hwbot.org/submission/3829104_

 

And, here's what the total system power draw is looking like with the modded 1080 Ti... (I no longer have GPU power sensor accuracy with the mod, so I have to look at the total system power to get a feel for the increase.) About 200W more than before. It is also a lot more temperature sensitive now. If it hits 52°C it locks up and has a TDR because the GPU does not downclock at all--stays fixed on whatever clock speed is set--as temperatures rise. Almost like having K-Boost enabled except that is does return to idle clocks with no 3D load, whereas it does not do that with K-Boost enabled. So, looks like a full PCB GPU water block or portable AC is needed to keep the GPU at 50°C or less once I start approaching 2.2GHz on core.

 

Does it TDR at 52 degC because it can't downclock then - as in NVidia normally downclock the GPU with increased temperatures to maintain stability? If that's so, then NVidia's downclocking driver behaviour with increased temperature is a sensible way to maintain a maximum stable overclock.

 

As best I can tell, that seems to be the case. I'm guessing to some extent, but it is predicable and I can replicate it. And, yes, the rationale makes sense to the extent it keeps a system stable. What doesn't make sense is the low temperature threshold.

 

Maybe it just proves that Pascal is very sensitive to temperature changes when it comes to maximum stable frequency for any given temperature & voltage. If so, then NVidia looked like they've done a good job with tuning their frequency/temperature curves, and perhaps the best way for us to overclock is let the driver do it's thing & lower the clocks, while overclocking to the highest stable frequency offset (while keeping GPU as cool as possible) - which is just the 'normal' overclocking process we all mostly do for Pascal. Unlocking TDP and unlocking more voltage (like you've done) is a positive though, although some GPUs don't seem to respond to more voltage, mine doesn't (1.062V max or 1.09V max it makes no difference to mine).

 

If you're stuck with a sucky vBIOS like almost all Pascal GPUs are except for the Kingpin and Strix XOC GPUs, there's not a good option and you have to do the just let it throttle the clocks thing like a broken piece of garbage, or keep it abnormally cold. If you have a vBIOS that isn't screwed up, then you don't have to worry about it as much. But, yes... Pascal is temperature-sensitive to a ridiculous fault... I think a lot more than it needs to be. You still have to keep it abnormally cool.

 

And Intel has started push similar Cancer boost for their brand new Unlocked Mobile i9-8950Hk. 50C is the target for the extra boost clocks Slightly better

 

Exactly. A meaningless spec that allows them to sell something they do not actually sell because nobody's laptop runs the CPU under load at 50°C. Maybe we cannot accuse them of being liars, but it is clearly a very deceptive practice.

 

Well ok, I'm looking at it more from a favourable angle towards NVidia - as in I think they've included small temperature downclocking in the driver to maintain stability while at the same time maintaining a maximum possible stable frequency, which is kind of supported by your observations that it crashes at the first downclock temperature point on your card when it's been prevented to downclock. In this light I see this driver behaviour as a positive, a way of maintaining maximum possible stable frequency. Either way, like you say, best to keep GPU as cool as possible.

 

I would bet that's more related to ensuring that thermal spikes don't exceed safe temperatures in notebooks that don't have good cooling. If they set the target significantly higher I could imagine the 'massive' boost clocks creating heat spikes close to 100 degC before the cooling system would have time to react, and thereby possibly causing damage or immediate thermal shutdowns or severe below base clock thermal throttling. But, to be clear, I'm not a fan of the whole thin & light disposable laptop brigade, laptop cooling systems need to be beefed up more!

 

Nice, just shy of x3 higher GPU score than me!

 

Which gpu did you have?

 

Zotac GTX 1070 AMP Edition, ha yep, not expecting it to match sli 1080ti, but just drawing a comparison.