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

Oh for sure, I think we all know who the provoker is around here lol...

Great score btw... That's awesome..

Looking forward to things getting cleaned up...

https://www.3dmark.com/fs/19011465



Edit: ah I just saw your edit haha... Yea it's awesome to see new heights all from a laptop! I hope to get back to the A51M after finishing up this desktop stuff....

 

Haha, nothing about that. I only referred from his post Good score. Hope all is well. Best greetings... +rep for the FS record

 

It always "worked" or rather 'not really', just like any other GPU without SLI connector.

THX!
I had to borrow an RTX 2080 and put it in this weekend to see what can be done...mobile RTX are all hard locked to 2100Mhz...not sure if I'll have enough time with it to fix all the 'issues', but it's a good start.

I know! Things just tend to get out of hand quickly if you make it about brands.
There are very few people left that know how to bench, so let's try not to drive new user away by arguing too much among ourselves...

 

speaking of benching, im almost done with dialing in the timings at ddr4-3300 on my new ram kit took quite a while to go through all primary, secondary AND tertiary timings one by one and test for stability *grinding mode activated*

ill be sure to post some performance numbers vs. spd speed, initial 3300 timings, tweaked 3300 timings and my previous corsair 3000 xmp settings

good thing about this: now that i know what this kit/mobo combo can do in terms of timings, i may be able to go further with the clocks, lets see how that pans out....stay tuned!

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

 

You say " nothing about that," but you need to realize that the way that you come off is very condescending and very conniving so perhaps it's best if you just not say anything at all if it's not positive. The goal is to have fun and to share the joys of benching with pushing these systems. Up until @Prema dialed in 27K on a laptop (which is fantastic), 26425 was the score to top so now that the bar has been lifted, you can bet guys who do bench are excited to reach for the new mark.

You don't even have a RTX system with a notebook card, yet all you do is roam the threads to find any means to trash talk any other systems than the brand that you own? It's really obvious what your intentions are so just stop. Really.. it's childish and completely unnecessary.

At this point the A51M has nothing to prove if it can bench well for what it has along with its known limits. It would be amazing to get Prema custom bios on there of course, but for what it's able to do stock it's pretty impressive. Have you forgotten what graphics score it is able to reach? 30700's+ in case you've forgotten. The CPU is stable at 5.2GHz as well and perhaps we might be able to see 5.3GHz stable enough for benches soon. It's still a new system and there's definitely more left in it for sure... when I get back to it after my other projects we'll see how much farther it can go.

I really enjoy this part of the forums, because it's a lot more mellow without all the trolling that exists over in the AW threads so don't ruin that by bringing that mess here. Whatever it is that you do in kissing up, trolling or whatever is your business, but I'm just asking you to be nice and to not provoke any brand wars, because most of us here other than you own several brands of all types, so if anything the fan-boyism that you have with clevo doesn't need to be reminded beyond what we can already see from your signature.

Play nice and act like an adult. Show some love and not hate. Like what Prema said, don't ruin it with your provoking in these threads... be nice and mature.

 

Right on! Keep us posted for sure. It's crazy to see a lot of RAM prices dropping so to see stuff like this with what you're doing in that RAM speed range is awesome. I have a pair of 3200 RAM sticks that I'll be using for the 2nd test bench so I'm curious how far those can be pushed. Cool stuff... looking forward to your results.

 

I don't know where I shall start and stop repling on your post. That you drag in that I don't even have a RTX system with a notebook card + don't show all my setups in my sig.... I don't know what to say. All I can say is re-read your own post #4797 as I said in previous reply. + this one. And yet you start all over again. I think we stop here.

 

Just leave it at that guys! You both have a point but you are talking about different topics.

I really do like the 51M beyond it's very existence as another DTR on the market and would love to push it, but without AW backing I don't see that happening.

 

Bullzoid mods production GPUs and makes them what they should have been if they had been designed properly... Well, maybe it is more accurate to say that he used to do that... when the mods actually worked.

I think that article tells us about the others that can run higher voltage just fine. There's a special reason for that. They didn't buy a GPU. A small group of individuals was chosen to use a special product that nobody else has access to, and doesn't accurately represent the product that is available for purchase. They may have earned the privilege of being chosen, but that doesn't belong in the same data bucket as the results of those that were not hand-picked to use an exotic product that nobody else has access to. Nothing wrong with being special. I'm happy for those that were chosen, but it doesn't do squat for the rest of us that are stuck with broken mass production trash.

Several overclockers, including Rauf, OGS, Rsannino, Phil, Ronaldo, and ZeR0_Dan were invited to the GALAX OC Lab last week. All of them were given the GALAX RTX 2080 Ti HOF OC LAB edition graphics card which is a custom designed RTX 2080 Ti graphics card that is built purely for Liquid Nitrogen overclocking.​

If I had to guess, they probably were not allowed to take these unique specimens home after using them at the special event that was crafted to make Galax GPUs appear to be something special. If they were allowed to keep them, that would be a really awesome gift for the few that were chosen.

LN2 wouldn't allow my GPU run like those did because it was crippled by design, and done in such a way that mods that used to effectively cure their problems are no longer effective. As such, there is nothing for the rest of us to aspire to and no goal to aim for because those results are not attainable with products available to the rest of us.

 

Bro, you know you could buy that card? If/when it comes back into stock. http://galaxstore.net/GALAX-GeForce-RTX-2080Ti-HOF-OC-Lab-WC-Edition_p_180.html

The question should be, does the 'retail' card contain the same firmware/BIOS that those oc'ers had access to?

 

Yes, firmware is a big deal. Huge deal. Perfect hardware is totally pointless and runs like garbage with cancer firmware. But, we don't actually know for certain that the hardware available for purchase (if and when it is ever actually available) is identical to those GPUs used at the event. But, to your point, even if they are identical the retail product will run like utter garbage with the firmware we have access to. The Galax HOF vBIOS is one of the worst I have tested on my EVGA 2080 Ti. The performance and benchmark scores are better with the stock EVGA firmware in spite of the over-hyped power limit on the Galax GPU being something special.

 

Is this the card you have?
https://xdevs.com/guide/evga_2080tixc/

 

No, but close enough. Mine is the XC2 iCX, which has a few more circuits on the PCB. That web page is where I got initial info for the shunt mod, which works great. However, the voltage mods do not work, as Bullzoid later discovered. In theory the voltage mod should work, but does not because NVIDIA added a ton of extra garbage circuitry on RTX cards, so the volt mod causes the GPU to malfunction and perform worse than stock.

Some of what they posted on the shunt mod is also wrong. I and others had to modify a couple of things because following that article exactly causes the GPU to be stuck in low power 2D mode (300 MHz). I think the person that did the article was operating on a few assumptions that Kepler, Maxwell, Pascal and Turing would all handle the mods exactly the same. And, that was true until Turing.

 

I was more wondering about this part...

Could be because of the Vbios, but that's speculations.

World-records on this model after the modifications and extreme cooling.
Vince “K|NGP|N” Lucido already had completed initial set of benchmarks, securing performance world records in next benchmarks:

• 3Dmark TimeSpy Extreme : 9312 points, RTX 2080 Ti card clocked at 2.4 GHz core / 8.7 GHz effective DDR6, cooled by liquid nitrogen and Kingpincooling.com ICON
• 3Dmark TimeSpy : 18925 points, RTX 2080 Ti card clocked at 2.4 GHz core / 8.7 GHz effective DDR6, cooled by liquid nitrogen and Kingpincooling.com ICON
• 3Dmark FireStrike Ultra : 11081 points, RTX 2080 Ti card clocked at 2.4 GHz core / 8.7 GHz effective DDR6, cooled by liquid nitrogen and Kingpincooling.com ICON
• 3Dmark FireStrike Extreme : 21421 points, RTX 2080 Ti card clocked at 2.4 GHz core / 8.7 GHz effective DDR6, cooled by liquid nitrogen and Kingpincooling.com ICON
• 3Dmark FireStrike : 38704 points, RTX 2080 Ti card clocked at 2.4 GHz core / 8.7 GHz effective DDR6, cooled by liquid nitrogen and Kingpincooling.com ICON

 

Yes, it could definitely be the way the mod interacts with the vBIOS. Vince may have had excellent results with those hard mods if he had a vBIOS that did not kick the GPU into crisis mode. Everyone with stock ASUS, EVGA, MSI and other production firmware have not been very fortunate. Shunt mod works, but voltage mods do not. If you shunt the resistor on the lower corner (same edge as the PCIe connector) that will also trigger crisis mode (300 MHz).

 

It's not just vince, it a quite a few others. That was just the first one I found. Most people looking to pass the 2200mhz range use high end Lighting/Galax/Kingpin cards Under LN2 as they are built differently. Cards based off the reference specs are a trial and error and hope for the best scenario. Then once you find one that will get there. No amount of added voltage is going to help if the card is not below in the serious -C underload class.

Exactly why serious overclockers do not by referenced boards....
And that card works as a normal card as well.

 

Buy them, or receive them? I guess we will find out if I can get a Kingpin 2080 Ti.

I didn't need one before. Kind of sad that I do now. Takes part of the fun out of the equation when mods don't work.

 

Most people "buy" them, it's only a handful that get them for free as samples. Or get them during benching competitions, but they do not get to keep them. I found that out at the last competition i was at in Vegas.
No, you always needed one, it's just that these high end cards only recently came out. So the spots were pretty much filled with referenced cards, but now that the big dawgs came out....all those scores get knocked off with anyone willing to spend the money to buy one. Many many many times I have watch my scores go from top 5 to 30 or lower in the hof ranks after the real cards hit the market. This time is no different unfortunately.

 

https://xdevs.com/guide/2080ti_kpe/#cbios

Kingpin vBIOS for those interested in seeing if it plays nice with other cards. I guess main draw would be V/F curve working properly with a much higher TDP. The Asus XOC was useless without the V/F curve. This could be more promising if it works.

@Mr. Fox @Johnksss

 

It'd be best if you just stop replying all together as it's clear that you don't get the point. That's exactly why you don't know what to say. Perhaps you should re-read my reply to you so that you can maybe understand the second time around.

 

I see you continue and can't stop. I end it here with to say... Take it for what it is. I endorse all my posts. All is correct written and right on target.

Spoiler: End from my part

Why the fire extinguisher was mentioned in my post above… Quality is much more than use of metal in inner chassis.
http://forum.notebookreview.com/thr...m-owners-lounge.826831/page-750#post-10895394

 

I think the only remedy to a limited item (commodity for us) is to keep pushing for proper categorization so that it is clearly distinguishable.

This is what I've been saying all along about 3DMark / hwbot and explaining that there are different classes that need to be recognized accordingly... That's only fair and practical... why this hasn't been done? That is what baffles me even more...

It terms of the cards themselves between KP and GALAX, I wonder if the hand picked Samsung chips come from the same pool or if they are entirely different (bin wise) .... either way we know they are both top picks, but I wonder.... hmm

.

 

Wow you just don't get it don't you. It's like you only want to see what you want to see and that's it... this is ridiculous and a waste of time. Why don't you just stop and take time to read and understand. Quite trolling.

Do everyone a favor and DM me if you'd like to talk and keep it off the boards. Respect the good nature of these threads on this side...

 

Alright bi*tches

Time to compare my previous Corsair 2x16 GB DDR4-3000 Kit with my current results for the G.Skill 4x8 GB DDR4-3800 Kit at 3300 Mhz. I included Stock, XMP, OC with stock timings and OC with tuned timings to the max.

Lets start with clocks, timings and voltages:



Look at what kinda timings the G.Skill kit is capable of! Even beats the Corsair XMP settings, crazy stuff. Made me hopeful for a nice bump in bandwidth and latency

Now let's take a look at the actual performance. Basis for this was the average value of 10 consecutive Aida64 runs to minimize any variance in results. Read/Write/Copy is given in MB/s and Latency in ns.



Crazy, right?! The G.Skill kit allows for higher bandwidth AND lower latency at the same time compared to the Corsair kit! Nice! This was also the very first time i saw more than 50.000 MB/s RAM bandwidth on a laptop.... That was in the Write category, btw.

Next step: Let's see if I can get 3400 Mhz stable, now that i know the limits of the G.Skill RAM timings at a given clock speed... Cuz if you set the timings too lax it'll also give you errors!

edit: ok im officially dumb (or just tired )

i botched the performance column above, the correct values should be:

G.Skill 2400 Stock = 100.00%
Corsair 2666 Stock = 106.84%
Corsair 3000 XMP = 119.24%
G.Skill 3300 Stock = 128.89%
G.Skill 3300 Tuned = 132.52%

sorry about that! the rest is correct though edited the picture above and removed the incorrect performance numbers to avoid confusion.

 

Very nice! I was looking forward to hearing back. RAM is a funny thing ain't it? That's what makes it so daunting to tune lol...but at the same time very rewarding... great work.

 

Great work Brother @jaybee83... +5 rep

 

YESSS! now with the new info on timing limits i got 3400 Mhz tm5 stable!

Next step: 3467 Mhz

 

Have you tried flashing it yet? I will later today. Just curious if you tried it and if you could actually flash it.

Did I miss something? New info on timing limits? Link?

 

haha no you didnt, i was referring to the tightest possible timings i could reach at a given frequency. thats what i got when tuning the timings at 3300 mhz. before that i only had the stock timings (too lax) and the xmp timings (too tight) to refer to, which didnt allow me to go beyond 3300 mhz stable. but now thanks to more available data i can go further

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

 

edited above post and corrected the performance values in %. sorry about the first incorrectly calculated values, guess im tired

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

 

Looks like you are having fun with that. Only question is when you have to loosen timings where performance degrades. AIDA64 only does peak readings. You'll need to do sisoft for overall average and run actual benchmarks to compare differences in performance.

Continue with what you are doing, find the stable timings for each frequency, then when you've reached the fastest frequency you can, go back and compare slower frequencies with tighter timings to higher frequencies with looser timings. Also, remember, even if stable, over tightening some timings can degrade performance at times. Keep an eye on that.

Also, the creator of the Ryzen DRAM calc does have plans, at some point, to create one for Intel. Unfortunately, Zen 2 optimizations is his next objective, so may be awhile before he gets to it.

Any who, let us know how your march to improved speeds and timings goes.

Sent from my SM-G900P using Tapatalk

 

After finishing going after maximum OC speed... if you want some guidance for primary & secondary & tertiary sub timings.. I might help you very limited info. related tweaks to laptop world.. all the information are always targeted for Desktops

 

I tested all 3 K|NGP|N vBIOSes and while they function, it appears the firmware accomplishes nothing on my 2080 Ti XC2 iCX card. The voltage is still capped at 1.093V (which means my max core OC is not any higher). Classified R2 app thinks the GPU is a K|INGP|N, but the sliders do not appear to change anything. (The Classified app will not launch without K|INGP|N firmware.) Understanding that the software tools may not recognize the higher core voltage, I tested it using Classified and the max core overclock is the same even when I adjust everything much higher with Classified. I also have VRel and VOps PerfCap reasons in GPU-Z that I do not have using the Aorus Extreme vBIOS.

But, it was definitely worth trying. Looks like the Green Goblin has perma-crippled voltage on everything but maybe a K|INGP|N 2080 Ti. Figures... bastards.

The only thing that seems to actually do anything is the LN2 firmware (which I have on it now) does make the voltage readings show as zero. I think that is intended. It is also supposed to remove thermal limit readings. Not sure if that works or not since I cannot overclock any higher due to capped voltage nonsense.

@Johnksss @Talon

 

Is there a reason why you're not raising vdimm above 1.20V? (yet?)

 

Basically zero benefit from the K|INGP|N firmware on this GPU. Same mickey mouse 1.093V voltage limit. The Aorus Extreme 366W power limit vBIOS produces exactly the same result as the K|INGP|N 520W vBIOS. Still showing same max power draw from the wall based on same core clock and voltage limit. But, no CPU-Z PerfCap reasons using my stock EVGA firmware or Aorus Extreme firmware. But, at least the K|NGP|N firmware (LN2 vBIOS) doesn't impair performance like the Galax HOF and other so-called "enthusiast" vBIOSes do.

 

K|ngp|n vBios and tools do not work as intended on anything but K|ngp|n cards. Starting from K|ngp|n GTX 780 to RTX. And vice versa.... One gets crappy results when trying to flash the other vbios files on it.
RTX app doesn't work on 1080 ti and earlier either. And different versions of the Classified app do not work on all versions of K|ngp|n cards either.


Edit:

I think the arguments in this thread pretty much relate to what's going on still today. Only they tighten down a bit more....
https://forums.evga.com/STATEMENT-B...LE-WITH-VOLTAGE-NO-MATTER-BRAND-m2362059.aspx

 

i actually start with 1.35 vdimm and vccio/sa both at 1.25, then tighten timings, then lower the voltages again. so far stock voltages are stable

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

 

Can you not get even higher Mhz stable with increased vdimm voltage though? 1.35V would allow for greater Mhz no?

 

in theory yes. but i havent gotten anything higher than 3400 mhz booting yet, even at 1.35V vdimm and 1.25V IO/SA. thats the thing with RAM, its not just about voltages, but also timings. if they are too tight it wont work, if they are too loose it wont work either. need to dial in the "golden" values to get the system booting and error-free...

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

 

Bottom line of testing Notebook RTX for a week:

There is no point in even owning them as an overclocker in their current state as
NVIDIA has introduced ''clock-block 3.0" by hard locking all of them to 2100Mhz.
At that speed I have 0.050V (50mV) of unused stock voltage left on my sample,
which can't be used for anything.
Power limit was just a distraction and has never been the real issue.
In the end every GPU will hit that same wall and reach the same results.
This time everyone can be a "winner" aka "Kindergarden Olympics":

https://www.3dmark.com/fs/19038559

Hooyah! (•̀ᴗ•́)و ̑̑



@Mr. Fox The throttle reason you see means that the vBIOS would allow your chip to run on higher voltage but that your specific GPU has reached it's ASIC limit at that voltage step.

Those KP cards are all cherry picked...

What you could try is to first set a curve that keeps the the chip from hitting that ASIC limit, confirm with GPU-Z, and then only use the OV slider a single step above to see if they implemented it in a way that actually only adds voltage on top and doesn't just change the default boost table/curve.
Keep in mind that software won't show you anything above 1.093v even on the KP, so you would have to bench to see where it actually is at afterwards.

@jaybee83 nice results.
On my own RAM I see better RL results with 3200Mhz at CL16-16-16 than 3400Mhz at CL18-18-18.

 

makes sense, 3200/16=200, whereas 3400/18=188.9, so 3200 CL16 should definitely be faster

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

 

Here's a better table to help compare cl in real ns for each frequency.
https://www.reddit.com/r/Amd/comments/63gc1s/ram_speed_and_cl_equivalance/

That does not take into account effects of other timings.

Sent from my SM-G900P using Tapatalk

 

Word! RTX clock block 3.0 from the half-cocked skoks with a crock full of crap.

Awesome to have your validation that power limit is a silly distraction. I've been saying that since I first got the RTX 2080 Ti shunt mod done and most people are not listening. Maybe they'll listen with both of us saying the same thing. The PerfCap reason showing power limit is triggered by the lack of voltage that is suppressing core clocks (which uses less power) not the actual power limit, so that is misleading. I've got a higher power limit than I can even begin to use due to the castrated voltage cap... crap... cap crap.

 

Stupid IMO to handicap the enthusiast community as a whole. They better pull their head out of their collective arses, if they want future business from the aforementioned community!

 

@Mr. Fox , I sent HWBOT a message on Twitter to see if they will restore my account. I gave them my username and password and email address.

 

hwbot just gave us Notebook RTX user the go. May the games begin!

 

Should be a pretty narrow point spread, since they are all locked to the same max core clocks. The only real variance will boil down to thermal management and CPU tuning. But, everyone gets a trophy for participating. We can likely expect some drama as they discover that bragging rights about having a higher "power limit" are basically meaningless.

 

This is why in glad I got out of laptops when they didn't allow upgrading my p770zm from Maxwell to Pascal. Too much cost for forced gimping when you can build a beast desktop for the same or less.

Sent from my SM-G900P using Tapatalk

 

Me, too. I haven't regretted it... at all. Laptops are so "not worth it" anymore. And, they just keep giving us more and more, and more, reasons to say that they're not worth it. Nonsense piled on top of more nonsense.

 

Kind of why I still have these in a box un opened yet. (For the shunt mod)

They did this to most of us, not all of us. More like 99 percent would be closer to the truth.

Depends on the true max draw. Stock clocks can max out 200W.