Just in case 
Overclocking "Raptor Lake refresh" - 14900k,14700k, 14600k etc results, bins and discussion
31 reading
Nizzen
6,061 - 6,080 of 23,546 Posts
Joined
·
6,329 Posts
Igor lab tested (199 Chips) average SP from all 13900KS was 108.1 with the worst being SP101. I thought I was way closer to the worst than the best.
Maybe we just have a very bad/ low SP sample lol. I don’t know that many people with worse than SP104 besides you of course. 😂 (Kidding)
I never cared that much though, my SP104 13900KS chip was amazing. 13900KS goes down in history as one of my FAV’s
Joined
·
223 Posts
Sorry forgot to mention, P 113, E 82
Joined
·
743 Posts
is it cinebench15/15e that requires like 20ish mv more than cinebench23 to pass for vmin? after finding this chips vmin with cb23 what else should I use to test for overall stability besides just tossing on some games/trying stuff out - i'm on an aio right now so not trying to oc too heavy or anything yet
and is cinebench24 even worth using at all? is it another that requires more or less mv compared to cb23?
and is cinebench24 even worth using at all? is it another that requires more or less mv compared to cb23?
Joined
·
656 Posts
Take all of IGORS binning results and THROW them out the window. It's not a realistic estimation by ANY means. Take it all with a grain of salt.
His company has cult like status and Intel knows it, and they also help him procure his testing samples. Remember, he's not actually going out and spending pocket money on the 200+ processors that he uses for these tests. He's getting FAR MORE "good to bad" then any average consumer would EVER get. He's almost being given cherry picked tray's and then basing his findings off that. Do the math, this tactic and built in mechanism acts like a driving force so that retail goes out and buy's a new upgrade. Just look at his 13900KS results... SP 101 the worst? I find that hard to believe considering I personally went out and purchased 20 separate 13900KS's on release day and at least 25% of them were SP 97 - SP 105, and 72- 75% of them were SP 106 -110, I only had 2 processors out of that sample purchase that landed above SP 110 Global. If you were to go off his results, I am either the most unlucky human being in the physical realm, or his tests are HEAVILY skewed in Intel's favor.
Additionally, go glance over his 14900K bin results. It's almost comical, since if you were to base your buying decisions off it alone, you would think you actually had a GREAT chance of getting an SP 100 or higher 14900K. If you actually purchased or binned any, you would know this is FAR, FAR , FAR FARRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR from the truth. Again, I purchased 20 processors in total. 18 from the US, and 2 from China. Over 90% of my processors were under SP 98. I only had 3 chips over SP 99 ( 99, 102, 103).
I have come to realize you cannot trust the review - what so ever- of any person or company that has any kind of funding occurring. Especially when the funding and or sponsorship comes from the manufacture of the product that you are reviewing. Imagine trusting a De8auer review of a Asus RTX 4090 where he gets 3100 on his GPU. Do you HONESTLY think your going to go out and buy an Asus Strix OC 4090 and get 3100mhz Clockspeed because you saw De8auer do it? no..... Because Asus cherry picked a sample intentionally for the review. The same thing with JayzTwoCents and his BS reviews and product demonstrations. KitGuru does the same damn thing when they review products. It's all BS.
Joined
·
6,329 Posts
That really sucks. And it does makes sense, I would not want to bin for a good 14900K though it’s wayyy too expensive and I couldn’t possibly afford buying 50 chips just to find 1 or 2 good ones.
I also suppose It would still indicate my SP104 KS is a lower bin chip if over 72% of your 13900KS chips were all SP106+ or higher though. That’s not too bad.
Joined
·
1,889 Posts
He does have a valid point though.
Joined
·
4,489 Posts
No he doesn't. Igor is not supplied his tray processors by Intel. He gets access to the CPUs from MIFCOM, a PC store. It took all of three minutes to find that information.
But nah, let's just slander him for being a paid-off shill instead.
Joined
·
2,112 Posts
Even if Intel helped him or whatever, bringing all the better chips that are already "KS" doesn't seem possible.
Joined
·
2,112 Posts
I've tried a total of 3 13900KS samples I got my first one from microcenter on launch day. it was x246 118/95 and the one I have rn is batch x304 119/89 from B&H(can't run VT3 8000)⚰. The last one was x249 batch irrc was 114/86 which I returned at B&H.
If his point was to spew garbage with zero actual evidence, then yes. Valid.
The limitations and how you should interpret the data is clearly layed out on the first page of the "article"
Parently his sample size of a poultry 20 units is somehow magically more relevant, a real knee slapper.
Joined
·
469 Posts
Didn't pass, or even close to it. UNLESS I am doing something wrong (first time testing this way). I stopped at 1.4 with it getting through like two runs or so. Will continue to test tomorrow.
I reset to optimized defaults, turn on MCE (no power limits), no XMP/OC'd memory, set "Actual VRM Core Voltage" to manual mode 1.34, sync all P-Cores to 58, all E to 45, turn ring down bin to disabled, set min and max cache to 48, and of course turn on LLC6. Am I missing anything or doing something incorrectly? If I am not, does that mean this CPU is trash?
Joined
·
2,112 Posts
What kind of cooling do you have? If it's an AIO it'll fail
Joined
·
13,203 Posts
Your post is confusing. You have two chips in your original post. Which one are you testing?
If you're using two different chips in the exact same board that are of completely different types (13900KS and 14900K, or 13900K and 14900K), this sometimes plays havoc with the V/F tables. However your V/F "seems" to be okay.
Passing R15 Extreme at 1.34v set + LLC6 usually requires god tier silicon. But this also depends on your temps. You should be able to pass it at 1.37v set however easy.
Ok so let's try to get back to the basics.
#1: are you running HWinfo64 "Sensors only" in the background so you can see if you get logged WHEA errors during these runs?
#2: What is the exact lowest possible "Bios set" (and load voltage--check the lowest reported vcore in HWinfo64--make sure C-states are disabled OR windows power plan is set to maximum performance or ultimate, to avoid your CPU downclocking to 800 mhz and reporting 0.7v vcore minimum) Vcore that you can do to pass three consecutive runs of Cinebench R15 Extreme (mod) without a WHEA error, app freeze or appcrash? What's the load voltage reporting as?
#3: Can you pass three consecutive loops at : 1.390v Bios set + Loadline Calibration: LLC6, x58 / x45 / x48?
And finally
#4:
What are the load temps (Core max, and CPU package) reporting in HWinfo64? (these are the highest of the individual core readings)?
Joined
·
2,112 Posts
I did actually. I'll post it
Joined
·
2,112 Posts
heres r6 benchmark HT on vs off. pretty much every game i know of does better with HT off and you can also clock higher without losing stability with HT off
Joined
·
469 Posts
Hey Falk thanks for the help on this.
Currently testing the 14900k /w your suggested 1.390, LLC6, C-States Off, 58/45/48.
I really, really should have tested my 13900KS in this manner prior, I still have it so may have to throw it back in.
Lowest "Bios Set" , sorry for the dumb question, where is that exactly? Is it the "Core Voltage" listed in the bios Hardware Monitor Summary? If so it's currently reading 1.368 V (with 1.39 set as actual VRM)
VCore under load is 1.270 in HWInfo
I was just able to pass 5+ R15 Extreme's at 1.390, I didn't disable c-states in my initial testing, could this have something to do with the difference?
My temps are good.... This is after 6 back to back runs of R15 Extreme. The thing is, ever since I swapped to the Apex Encore (from a Z690 Hero) I have noticed a decrease in temps, to the point where I am wondering if something is wrong. I only have a EK Nucleus 360 AIO, mind you the ambient in the room is 18c, so it's a pretty cool room. But even at a 310w power draw I am in the low 80's, I could be over thinking this though.
Joined
·
469 Posts
Ok now I don't know what is going on. I just did 6 passes of R15 Extreme with 1.34 58/45/48.....
I did flash an older bios and back to 0071 on the Apex Encore, as you kind of alluded to maybe the VID table was bugged, but it remained the same. So that and disabling c-states.
Vcore in BIOS = 1.314v
VCore in HWInfo = 1.225v
CPU Package Power 292w
Ok one other question, so maybe these results are invalid, but it's late and my brain is starting to shut down lol. I am noticing when running these tests my CB23 score is much lower than normal (41k ish, but when running these Vmin at 48/45/48 it is about 37k). Is that to be expected?
I did flash an older bios and back to 0071 on the Apex Encore, as you kind of alluded to maybe the VID table was bugged, but it remained the same. So that and disabling c-states.
Vcore in BIOS = 1.314v
VCore in HWInfo = 1.225v
CPU Package Power 292w
Ok one other question, so maybe these results are invalid, but it's late and my brain is starting to shut down lol. I am noticing when running these tests my CB23 score is much lower than normal (41k ish, but when running these Vmin at 48/45/48 it is about 37k). Is that to be expected?
Joined
·
13,203 Posts
No, 48/45/48 should give you about 42,500 points in Cinebench R23, if "EC" support is disabled in HWinfo settings. If it's enabled, slightly less, maybe 500 points lower, but 37k is very low.
I "think" R15 score should be about 1730 or something. But I forgot.
You said that you are running JEDEC memory (not XMP?).
6,061 - 6,080 of 23,546 Posts
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
