6.9 GHz on All 8 Cores: How I Set a World Record with a Hand-Picked Core i9-9900KS

Exploding PSU

Honorable
Jul 17, 2018
461
147
10,870
This is one of those article where I say "I'll never be this cool" (heh). Well done sir, that was a blast to read.

Anyway, Intel sure love the 7.3 GHz magic number. I remember back when 8086K had just launched someone in Anandtech also reached 7.3 singlecore on a retail chip (I think it was lucky_n00b, probably, my memory is rather weak) while reviewing it. For some reason when I read 7.3 here that 8086K came to mind.
 

Colif

Win 11 Master
Moderator
Crysis was written at a time when people thought CPU speeds would be 10ghz just a few years after release. That was 2007, we been stuck at 5ghz ever since as they need a new substance to make CPU out of before they can break 10ghz on air.

Crysis can still make good PC struggle now, simply as it assumed we would have more than we got.

so its why i still only have a top speed of 4ghz as I don't see point in going max speed and reducing the cool factor of next PC.
 

CerianK

Distinguished
Nov 7, 2008
260
50
18,870
This specially-binned chip got as quick as 6.95 MHz on all of its cores ...
I thought it would be about 1000x faster (i.e. should be GHz).
Crysis was written at a time when people thought CPU speeds would be 10ghz just a few years after release. That was 2007, we been stuck at 5ghz ever since...
IPC (instructions per clock) is 3x-4x higher since then, so GHz is usually not as important... but you likely won't see those improvements in Crysis, as it was apparently not optimized to take best advantage... so in this case, yes, GHz can be important.
 

bit_user

Polypheme
Ambassador
Thanks for posting.

Normally, flattening the IHS lowers temperatures by as much as 4-5C on the warmest core, as well as resulting in a more even spread of temperature between the dies.
Thanks for that. I have lapped my last few CPUs and heatsinks. I didn't do a before/after comparison, though. I just figured that since I'm cooling on air, I'll take every other advantage I can get.

Anyway, I probably stay on the coarse grits too long, but I often end up going until only copper is showing (except at the edges). That's just how long it takes to work my way through all the grits I use.
 
"Every 9900KS is rated to hit 5-GHz on all eight of its cores "

uhh.....this was how it was for even 9900k.....never heard anyone with one that didnt do 5ghz...

i know high binned ones let you go further, but stating this new chip can have 5ghz all core as a sellign point when old one didthat too....is cheap :|
 

larkspur

Distinguished
LOL, he used Team Group memory on an ASRock board.
https://www.tomshardware.com/reviews/t-force-vulcan-z-ddr4-3200-c16-memory-kit,6163.html
that choice could be the result of spending too much time on CPU optimization.
Thomas - If I was you I'd write up a whole separate article about it. I didn't originally read the article you linked until just now (There are soooo many different memory kits!) and am now in the know regarding these anomalies. But I definitely wouldn't have known that unless I just read your comment here. Anyway, thanks for pointing that out, and I hope you get the time to investigate this further!
 

Crashman

Polypheme
Former Staff
Thomas - If I was you I'd write up a whole separate article about it. I didn't originally read the article you linked until just now (There are soooo many different memory kits!) and am now in the know regarding these anomalies. But I definitely wouldn't have known that unless I just read your comment here. Anyway, thanks for pointing that out, and I hope you get the time to investigate this further!
I spent quite a bit of time and spread much of that work across a variety of articles, but the most startling finding was that Gigabyte boards matched badly to Aorus memory. I postulated that Aorus memory may have been developed by Teamgroup, even if it's being manufactured by Gigabyte.

From all my testing, I've found the if you want the best performance across all boards, use Adata. G.Skill is pretty close to Adata for that consistency, close enough that we use it in motherboard reviews. Crucial and Kingston are close to G.Skill for consistency across motherboard brands.

I'm not certain if any of the test you do are impacted by the advanced timing oddities of Teamgroup memory. One more thing I should note though is that this doesn't seam to have much impact on AMD boards, mostly Intel.
 
  • Like
Reactions: larkspur

CoD611

Commendable
Dec 10, 2019
4
1
1,515
Does this really count as any sort of record? The 9900KS is simply a binned 9900K. Which, a long time ago, had der8auer hit a 7.6GHz overclock on all cores.
 
Dec 6, 2019
2
0
10
From the thread I linked.
"Soo it looks like he did beat the record, by one point, and the screenshot they included the article is showing the wrong score. This is the actual new record: https://hwbot.org/submission/4278190...9900kf_3143_cb

Also he used a 9900KF not a KS. "

I'm just trying to figure out what is what. Post author indicates that the CPU used was a KF not a KS? or is this a different benchmark submitted?
 

Latest posts