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

Still a nice looking board. As long as it works, I suppose all is well.

Hey, what voltage does your SP94 11900K send stock at 4.8Ghz all core load during Cinebench around 1.225V?

 

I agree that having the newest technology is not necessary for it to be fun. If you're dealing with an extraordinary product it may not even be necessary to have the newest to be the fastest. We saw examples of that with X299 HEDT for several years. In spite of some things that I don't like about it, Intel 12th Gen performance is very impressive and there is no way a 10900K will perform at the same level even with a golden sample on LN2. However, the price tag associated with the upgrade is unattractive. And on top of that, the more difficult it becomes to install Windows 7 the more my interest in a product wanes. I have been using Windows 10 since it was in beta, just as I did Windows 8. It has been almost 7 years since it was released, but I still hate Windows 10 and view it as a garbage product.

It's because the Redmond Retards screwed up the RTC when they vomited Windows 8 onto the world and haven't cared about their blunder enough to fix the problem. Benchmark scores are sometimes artificially inflated or overstated because of it. Probably intentional to some degree. Nobody likes to be told their baby is ugly, even when it's true.

 

1.214v IIRC. The most dog of a 11900k I had pulled 1.284 at stock for CB23. That was the one that pulled 368w @ 5.2 The best I've seen was 1.194v which was SP103 I believe.

 

Wow! Hey, mine is 1.195V with auto voltage.it may actually be 1.194. Let me check.

 

That is pretty righteous but it would depend on the entire V/F curve. I also saw an SP83 have a vid around 1.197 too at stock then ramp up to match its SP83 rating.

 

How would I check this? It’s running stock right now, and I can run a Cinebench and it locks on to that 1.195V on all cores underload. Temps are 39-41C and the power draw is 140-144 watts while memory is at 4000.CL14 Gear (1) with 1.385V to IO/SA.


Before the delid the stock power draw was 148 watts. It really didn’t help it too much at all.

Unlike my other 11900K that went from 190’s to 170’s after delid.

 

That's the problem is delidded and running direct die on that MO-RA makes temps and pull suspect for accuracy but pull is pull in the end for your particular setup.

I think I mentioned it before but just leave it at auto and just bin up by 1 and see what it pulls at each bin till it gives up the ghost but running DD with that cooling and MB variance makes it hard to really lock down its ranking. What we DO know is it is definitely an above average chip.

Also, I think I asked before but what is the V/F curve up to 4.8 look like in XTU? It will show you the hardwired V/F curve up to that point. Then you can get an idea of how your environment is effecting the chip in regards to pull compared to the hardwired value and how your setup helps.

In the end, you won't ever really know till you pop it into an Asus board to get an SP rating.

 

Your SP94 on a AIO would smoke my other 11900K in low power consumption and low voltage lol. It was really embarrassing to say the least haha. Because, my 11900K was running on the best cooling parts, and it was just a joke, and your AIO cooled SP94 11900K still ran cooler with less power consumption than me.


One thing interesting is, my new 11900K is 148 watts stock before the delid. With less power consumption than your SP94 11900K. The delid did not reduce stock power consumption on my new chip. That is still within margin of error of 144-148 watts after the delid. With 1.195V with auto voltages stock or now, I would think it’s probably have to be a SP94 or potentially even higher. But I guess there is more to it than that, and I’m speculating.

I wish I knew for sure. Sucks that only Asus tell us this information. It’s literally killing me, I wanna know so baddddd!!! Ughhh. Lol.

 

If you want at least an idea of baseline V/F just load up XTU and it will tell you up to the 4.8 wired V/F point. Compare that to what you get on your EVGA Dark and you will get an idea of how much the Dark is able to dial in and compensate. Then factor in the MO-RA and finally DD.

The problem is the baseline vid isn't the be all of all. That is why I gave an example of an SP83 with a 4.8 run like yours (1.197 vs. 1.194) at 4.8 stock CB23 runs. The fact yours won't do 5.5 all core stable even DD with a MO-RA is why I think it falls in the SP90-99 range. Until you slap it into an Asus you won't know the full, calculated V/F curve and where it ranks.

I know those SP100+ 11900k's when given a MO-RA and *sometimes* DD will do 5.5.

What we DO know is your current chip is heads and tails better than your previous chip. That is a product of your previous chip being garbage and your current chip clearly being above average. How much above average? I'm curious too.

I'm equally curious to see what my 10900k in my X170SM spits back too.

Load up XTU, have it grab your V/F and post a shot to see what it says.

 

Do you know anyone close by with an ASUS ROG Z490 or Z590 mobo? If they have the latest BIOS the Z490 supports 11900K. You can get your SP rating, F12 to screenshot it, then put back their original CPU. Had I known you wanted to know this bad I could have installed it long enough to check for you, but then it would not have been brand new.

The alternative would be to buy the least expensive ROG Strix mobo as an upgrade for your wife's computer and use that to check the SP rating for both processors. Unless that would be a downgrade for her system.

 

Any Asus ROG board? Like a Hero or Strix Z490 or Z590 board right? Even the Mini ITX ROG Strix?


This is a great idea. This Aorus board kinda sucks anyways anyways. And, I could swap It out for an Asus. Ok, I’ve got an excuse now! Cool. Thanks Lol.

 

Yes, even the least expensive ROG motherboard in the Strix line. I have a Strix Z490-H Gaming. It is a gamerboy mid-range meh product that is made to look nice with the red LEDs turned off. I bought it as an open-box return at NewEgg for like $135. (I can't remember the exact price, but it was like $100 less than new retail.) You can probably get a similar deal at NewEgg or Amazon. If she is using a GPU that cannot benefit from PCIe 4.0 there is probably no benefit to paying extra for Z590. Honestly, I don't think PCIe 4.0 even matters with the 3090 KP. Like a big to do about nothing as far as GPU performance is concerned... worthless IMHO. My Z490 Dark was a better product than the Z590 Dark, and that is probably true for most other top tier enthusiast motherboards with other brands. Newer keeps getting newer, but I am not seeing any indication of better. Higher prices, lower quality... the way of the future.

I wasn't impressed with the Gigabutt Aorus Z490 Master board that I had. It was an unimpressive piece of junk and I RMA'd it. I'd be apprehensive about ever buying another one. It was nothing to write home about, that's for sure. Although, I will say, it was not as unimpressive as the ASROCK Chai Tea mobo that I had at one point. That was a very overrated chintzy turd mobo.

 

That is what I’m effectively doing. The idea is to replace the aorus POS with the Asus z590 hero.

if you have the opportunity ditch the gigafail and get an asus for immediate SPs on 10th and 11th.

 

Z490 - https://www.bestbuy.com/site/asus-r...tel-motherboard-black/6416651.p?skuId=6416651
Z490 - https://www.ebay.com/itm/275044140731
I am not a fan of white, but if that is your wife's thing...
Z590 - https://www.ebay.com/itm/255249937219

 

@electrosoft

I downloaded XTU. Here is my VF curve.

 

That is awesome the Dark tunes and dials in almost exactly what the curve is internally. EVGA was late to the game but spot on with the Dark IMHO.

Go ahead and just bin up and run CB23 and map your V/F curve and see what it spits back to 5.3 at least.

I just installed the Hero in my test case. Motherboard works fine. Amazon Warehouse for the win first time in quite awhile. Description was completely wrong.



1.217 under CB23 pull @ stock. That sub 1.2 on yours makes it a pretty special laptop CPU at the least. I'm going to predict an SP98.

Asus Z490 said SP94. Z590 says SP93 on Bios 707 (Wall-E says hi!):

 

Ok so, let me make sure I have this right. Apply x53 in the bios, and launch XTU and then run the XTU VFCurve with auto voltage @5.3 ?


Yeah the Z590 Dark was late. But a great board.

Also, @Mr. Fox Evga just dropped a brand new 1.07 bios. Maybe give it a shot and see if that fixes your throttle issues with the 10850K.

 

No, XTU won't map past stock but seeing as the reported pull on auto matches the internal V/F curve, just bin up in BIOS and run CB23 and see what it pulls at each frequency (4.9, 5.0, 5.1, 5.2, 5.3) and that will give you a rough V/F curve for comparison.

 

I have had it for a few days. It did not fix it. All it did was installed a newer Intel ME for "vulnerability" fixes, which I have disabled in the BIOS anyhow.

 

Testing 1.07.

 

Anyone know why my read bandwidth is so low?

My write and copy are both higher. It’s kind of confusing.


I was thinking maybe 63-64GB read would be minimum with 4000 Gear (1) CL14-14-14-30-2T. I could actually obtain higher bandwidth with single rank sticks.

 

Some people have received their Optimus Kingpin blocks. One user(s) install is NOT going so well. His 3090KP PCB is looking like a banana. Just an FYI! Optimus is saying their block is very precise, and the install is something to take very seriously.


Here is what they said:




“Guys, please don't use a torque driver at any strength on this block anywhere. You will run into problems, please trust us on this. A socket like 3647, TR, etc has a much different install torque requirement than the backplate involved. 1.5Nm (the TR wrench) will absolutely wreck the block”

“Will update shortly with instructions for HM, but, again, please trust us that we've not only designed and built the block, but we've taken them and reassembled the KP blocks dozens of times and the other blocks hundreds of times. We know what needs to be done “

“Thanks yes, please follow our soon-to-be updated instructions “

“Interestingly, 1.5Nm (13.3ftlbs) will destroy the gpu. 1.5 will actually cause LGA 1200 to break the socket. We've tested it because we wanted to see if we included the TR thumb nuts for our TR block on LGA if people could use the TR 1.5Nm wrench. And the answer is no. We have piles of dead boards. And dead GPUs because we tested all this, pushing it to the limits”

“Another point: the speculation and over-tightening of the screw is what is causing HM's problems, while good ideas, trying to treat the KP block like a "normal" block is our internal concern. The KP is a very precise block, and people should really ask us questions about how it works, before trying to figure out DIY solutions”

“Here, the backplates screws need to go in so they catch and are snug, but are not tight anywhere near a CPU tighteness. Just slightly tight, prob less than 1ftlbs.”

“ONLY the die screws themselves should be fully tight”

 

6000Mhz CL36 arrived a day late unfortunately. Booted XMP without issue, turned it up to 6400Mhz and default XMP timings and no issues yet. Need to manually tune to really get latency down and increase the bandwidth. This stuff is already 100x better than that crap I had before.

 

That looks great!! Already 100GB+ bandwidth, and 56ns is pretty amazing considering it is not really tuned out yet.

Did you ever run Rocketlake? Did you run Gear 1 or Gear 2?

I can produce some amazing Gear 2 performance. Like 85GB bandwidth, and 40NS. But it seems like Gear 1 makes it all look silly in actual gaming performance. And it’s really confusing to me.

I have seen Alderlake show some similar situations. Is that still the case? Is 1:1 Alderlake better in BF2042, VS. 1:2 memory?


Seeing is how you play BF2042, and you’re after the best possible performance. I would like your opinions on that.

 

Finally got the 12900k and Z690 installed...SP81 Ugh:



First dirty run (Boot on old W10 install after BIOS update no tweaks or adjustments):



In case you can't tell, I just put a minimal amount of thermal paste on just to make sure she booted. Going to repaste and remove the little O rings on the top side of the motherboard for a touch more pressure then install my Strix 3080 and then backup and give a clean install since this is the 4th motherboard this poor OS has had to endure the last few days and this one has zero Z690 drivers installed anyhow then zero in on the IMC to at least hit 4000 1:1.

@Talon let me see the V/F curve on your SP91 for comparison to see how sad this chip is overall.

 

Great CPU overall! However, that is unfortunate that it’s not exactly what you wanted. Are you going to keep trying for a better sample?

 

Check out my beast mode bandwidth.


The BEAST is ripping!!! Not bad for a processor with a little laptop IMC lol.

 

O ring removal from 1700 kit and new spread application vs pea drop. Those drops in temps are most proper:

Before:



After:

 

I'm not sure. I have 45 days to return it since I'm an Elite Plus member so I'll hold onto it and work with it. Right now I don't know its limitations or have collected enough internal and external data to quantify and qualify its status for my use.
I may try another one but with some of them clocking in at SP60's to SP100's it is as 50/50 chance as it gets on another one.

So many reports of 12th gen "zomg overheating!" at stock and that's clearly not the case even pulling 240w.

As it stands now I'll probably return the Z590 Hero if I don't keep it as a 10th/11th gen SP testing/backup board since I bought an X170KM to go along with my X170SM that will be here next week so the 11900k is going to end up in there.
I could always swap it into the wife's system and sell her MSI too.

 

V/F Curve. Seems to diverge once it clocks up compared to yours.

 

Looking at the prediction table, I might have a cache winner and the P-Cores aren't too bad but the e-cores are going to be rubbish.
VID 1.199 vs 1.225 isn't the end of the world. That V/F +.076 @ 5.2 is a dagger to the silicon heart a bit. Hope mine isn't a bleeder.

 

This thing right here is doing some incredible stuff.

AKA

“Destructor of Frame-Rates”

DDR4 4000 CL14-14-14.

 

What does yours say for each score under AI Features?

 

Garbage P-Cores = 90, E-Cores = 65.
I'm testing the IMC now (currently @ 3600) and I'll continue to climb tonight and tomorrow to see if there is some redemption but I can almost guarantee this one won't make it to 2022 before being returned.

 

This Strix-A board? I think there are some newer BIOS floating around on the OCN forums, but not sure which one is best. I think I saw some people having issues with 4x8gb configs though. Not sure what you're testing.

I feel your pain in rollin' the dice hoping for a decent pull. Now we also have to worry about IMC being a huge factor in performance.

 

That is crazy impressive latency for an 11th gen beast. I still really love 11900K. The IPC gains, the fast single core, the way the chip comes alive once you actually tune it's IMC. The EVGA board really does seem like it was worth the wait for that gen. To answer your question about BF2042, I see no reason why you would gain much if anything over that heavily tuned 11900K you have there. Also DDR5 only runs in Gear 2 on 12th gen. I tried to see if I could force or trick it into going into gear 1, but nada. I feel like we need around 7000+ Mhz DDR5 to really have low latency DDR5 in Gear 2 and crazy throughput.

 

I'm testing 2x8 B-Die and 2x16 Sk Blahsticks that tap out at 3800 even with looser timings. I saw users were having problems with 4x8. I'll eventually pop in all 4. I really do not want to invest in any expensive DDR4 2x16GB sticks at this time while I wait for the DDR5 scene to clear up. If I manage to sell the 4x8's for a decent price I would probably just pour that right back into a set of much better set of 2x16GB sticks.

This one cost me $560 when all was said and done from Best Buy with a 10% off coupon and rewards. I know I ended up losing over $800 buying and selling 11900k's. It was fun though so I'm tempted to ride that train again. Your SP91 is a gooder not only because it's an SP91 but moreso
it has a high P-Core SP. That overall score is only dragged down by the e-cores but for me that would be the optimal scenario.

I'm still going to give this SP81 a fair shake and see what it has in the tank though.

The Dark is a sleeper G1 monster board for 11th gen. With @tps3443 run of bad luck (ish) with the 10th and 11th he had before and such a positive outlook still this is Silicon Karma balancing everything out with a monster 11900k and a pro delid.

 

@electrosoft I had the same memory as you, G.Skill 4x8GB 4000CL15 @1.5V xmp. Only they were super gold and bling like. Or as @Mr.Fox would say, they looked like “NY Pimp teeth” well, I sold my set for $250 ish used. And I bought my Corsair Dominator Platinum 2x16GB 3600CL14 set for $203. And my Corsair memory is actually better. And, I thought that would be a hard feat to manage.

You could easily snag $200 for your set, and grab the same kit I did or something similar. You’d increase performance due to dual rank. And get some really nice memory in just two sticks. One thing as well, the Corsair sticks run super cool compared to my Royal Z sticks. These sticks are very heavy feeling, and the cooling is something well designed. I can see why Corsair has not really changed them. Because they’re perfect. I always thought these sticks were overrated game boy stuff, that wasn’t all the well binned. But, I miss judged them. And, for $200 bucks it is an incredible value for 32GB of extreme DR overclocking memory.


Unless you are going to invest in a nice DDR5 set anytime soon. But either way, you wouldn’t really lose out on anything. You could certainly break even.

 

How NOT to cool a 12900k...... (I'm looking at you Origin). MSI boards always seem to pull the most at stock versus other boards.

In HWU's round up of Z690 value boards, yet again the MSI was pulling the most, Gigabyte was below them but still pulling too much and finally Asus boards were pulling the least.

Using auto settings as always, I found the same thing with the Z590 boards. MSI just pulling way too much followed by Gigabyte and finally Asus pulling the closest to 4.8 V/F point (1.217 pull vs 1.214 native).

Either that or I'd also like to keep at least half the kit for testing since they are pretty rock solid and a tested, known quantity when pushing IMCs for raw baseline.

With loose timings, how far can you push G1 or more importantly G2 with those DPlatz to get an idea how far they can stretch clock wise?

Did you order an Asus board to test the SP? I'm super curious to see what it is on that 11900k (both actually)!

 

I don't understand how these reviewers keep getting these temps. I've tested 4 12900K, and all of them have maxed out around 75-80c out of box settings max under load. Once you turn up the fans, and tune the voltage you end up with around 60-65c in CB23 with slight overclocks.

 

Yep, your G.Skill set is valuable being (4) high performance low latency matched sticks though. I know a lot of memory won’t run all (4) sticks at 4000CL15 and actually be stable. Especially with HCi Memtest etc. on like Threadripper, X299 etc. If you sale just 2x8 you’d lose some value there.


As for my IMC it does seem to give up the ghost around 4112 ish range with CL15. I jacked up the timings and saw the same issues again with it not posting around 4112Mhz. I can dial in a nice 4,100Mhz Gear 1. But I have not truly tested its stability with HCI memtest. I am most likely going to tune out a nice (Gear 1) 4,067Mhz tonight or tomorrow. Or 4,100Mhz (If it’s actually stable stable) One thing interesting is, I only send 1.300-1.400V to IO/SA which shows it’s IMC is not struggling like my other 11900K.


I have not ordered an Asus board yet. But I will soon. I’m thinking probably SP95, maybe slightly better.


I don’t even consider 5.3Ghz as a speed just because how easily 5.4 is accomplished though (Which is incredible to say the least) I have never owned a chip that runs 5.4 all cores so easily without even breaking a sweat. It’s really a set it and forget it type deal.

 

My guess is that they're using crappy cooling, crappy thermal paste, have poor mounting, or a combination of the 3.

 

Yeah this one is definitely going back. G1 IMC craps out at 3867. Errors at 3900 even at 1.5 and 4000 routinely gets the green light and won't boot. Even getting it to boot sporadically takes 1.55.

Boots on auto right up to 3867 no problems.

Switched to G2 just to make sure the memory or subsystems aren't the problem and it booted right up no problem on auto.

Popped the dual rank sticks in just to see and it won't make it past 3733.

 

Very unfortunate the IMC is not very good. I bet 12th Gen would be a ripper with some Gear (1) 4300. I hope you find a good one soon. I saw some numbers some where of Gear (1) CL16 3700 eating up DDR5 6400 in gaming. I wish I could find them.


Gear (1) is very interesting how it responds on 12th Gen. it really likes it I know that. So if you find a good one, dual rank DDR4 would be straight beastly on it.

 

So very true. I’ll hold onto it till I find another so I don’t have to break down my desktop again. I’ll at least see where OCTVB stands on it for my log files.

 

@electrosoft

Not sure if you've seen this. https://www.overclock.net/threads/o...lts-bins-and-discussion.1794706/post-28909394

Guy binned 12 12900Ks and his results.

 

I took the 11900K out again. And I put the rotary polisher right on the silicon die. Smoothed it down a bit, making sure it was perfect, there was actually solder still on the corners. So, I polished it up, and then I re-did the liquid metal.

My favorite setup. Direct die, without the die frame, and without the socket ILM on the motherboard.

The socket pins work as suspension to guide the die perfectly level to the waterblock cold plate, even if your block mount is not perfect. That die to block mount will be level


When you use a die frame with screws, the screws can harm the core to core temps. Because the die frame is not perfectly even at each corner, and that variance is going to be in the die to Waterblock contact.


Using no die frame removes any uneven waterblock mounting as a variable all together.

Check out these temps at 5.4Ghz during a Cinebench R15 run!!!

52-59C at 5.4Ghz all Cores.. wha huh?

 

So, if you know the batch number and are not ordering online, you could potentially cull silicon trash by simply looking at the label on the box? Of course, that is assuming you have something you can actually rely on as to what batch numbers are good and what batch numbers are trash, like that person in the link that tested a group of processors from various batches.

 

I think I remember reading an Intel rep commenting on that theory a few years ago when the 8700K dropped and people were trying to bin based on batch #s online. They said it doesn't work that way. However, when I got that 10900K SP103 chip I sent your way, I also got another chip from Newegg the same day. It had the same batch # or like 1# apart and it was also a very good chip. I think it was a low 90s SP which is still excellent quality. Of course that is all anecdotal soo.

 

Pretty interesting that he purchased so many. That 12900K SP103 looks fun for sure!!

It’s crazy how we have so many variables with silicon now. I feel so lucky with my current chip, it’s like the silicon is a G! The IMC is a G! Lol. But man, 12th Gen? We must get a lucky overall SP between P cores and E cores, and then we can hope on a good IMC as well.


It’s going to be an interesting journey hunting for a good one.