i9 9900k oc suggestions

[zayah]

Fairly new to overclocking and been running my i9 on stock for most of its time. lately been getting into it and been running my i9 9900k on 5ghz, 0avx, 45 cache and 1.27v thermals seem pretty good im happy that it runs fairly cool. on prime95 small fft hottest core is 86c and coolest core 67c. Im wondering if i should up the voltage and up the thermals for higher cache maybe aim for 47. Wasnt able to run 46 cache on prime with 1.28v. Idk if its worth to sacrifice the thermals for higher cache,  are the benefits worth it?

 

mobo: rog strix asus z390f   

[RONOTHAN##]

5GHz 1.27V is a really solid CPU. 

 

Realistically, you won't notice the difference, but at the same time, I would probably do that. There's not much harm in running a 9900K at 95C in Prime95, so I'd probably aim for 5.1GHz core 4.7GHz cache 1.36-1.4V if you can cool it, just for the benchmarks. 

[zayah]

28 minutes ago, RONOTHAN## said:

5GHz 1.27V is a really solid CPU. 

 

Realistically, you won't notice the difference, but at the same time, I would probably do that. There's not much harm in running a 9900K at 95C in Prime95, so I'd probably aim for 5.1GHz core 4.7GHz cache 1.36-1.4V if you can cool it, just for the benchmarks. 

 

I havent even tried 5.1GHz on the core yet but when i used to run auto settings it would pump up voltages to 1.35-1.37 and it would be on the low 90-95c on 5Ghz on prime95. I could try to see how much voltages it would require to even run 5.1Ghz and then check the cache but if its too much on 5.1 Ghz I think id have to add too much for the cache since I instantly bsod on 4.6 cache and 1.28v on 5Ghz but tbf I may have added way too little voltages. Im pretty happy with my OC right now but I was thinking if it makes sense to try to squeeze a little bit more and sacrifice the temperatures, I think my timing to getting into OC'ing is pretty bad since its summer just gonna make my room hotter haha.

[RONOTHAN##]

5 minutes ago, zayah said:

I think my timing to getting into OC'ing is pretty bad since its summer just gonna make my room hotter haha.

If it's that hot in your room, I would probably keep it where it is now, then overclock in the fall/winter. 

 

You could start diverting your attention to RAM overclocking instead, that does have a pretty decent performance impact. 

[zayah]

4 minutes ago, RONOTHAN## said:

If it's that hot in your room, I would probably keep it where it is now, then overclock in the fall/winter. 

 

You could start diverting your attention to RAM overclocking instead, that does have a pretty decent performance impact. 

Its not like super super hot but It is noticeable especially when running these tests. Was thinking abou RAM overclocking since im running 3200mhz cl18 4x8gb kit rn, but it seems so hard and like detailed so I might not even attempt. 

[RONOTHAN##]

1 minute ago, zayah said:

Its not like super super hot but It is noticeable especially when running these tests. Was thinking abou RAM overclocking since im running 3200mhz cl18 4x8gb kit rn, but it seems so hard and like detailed so I might not even attempt. 

Fair enough, 3200MHz CL18 is a very mediocre memory bin, so you probably wouldn't get very far. It is the harder form over overclocking, but it's not as hard as you might think. You set the voltage to the maximum safe voltage for the particular memory IC you've got without active cooling (usually somewhere between 1.4V and 1.5V, though there are a couple like Micron Rev. A and Samsung C die that don't want to go above 1.35V), loosen out the timings and go for the maximum, stable frequency. Then tighten up the timings until it's not stable. 

 

It's tedious, but it's not that hard. 

[zayah]

46 minutes ago, RONOTHAN## said:

Fair enough, 3200MHz CL18 is a very mediocre memory bin, so you probably wouldn't get very far. It is the harder form over overclocking, but it's not as hard as you might think. You set the voltage to the maximum safe voltage for the particular memory IC you've got without active cooling (usually somewhere between 1.4V and 1.5V, though there are a couple like Micron Rev. A and Samsung C die that don't want to go above 1.35V), loosen out the timings and go for the maximum, stable frequency. Then tighten up the timings until it's not stable. 

 

It's tedious, but it's not that hard. 

Tedious was the word i was looking for but I couldnt think of it, I think ill stick to my ram for now and might just buy a better kit at somepoint hopefully ddr4 prices come more down with ddr5 coming hopefully a bit more relevant with the new processors. Thank you for all the help and suggestions, appreciate it a ton!

[RONOTHAN##]

4 minutes ago, zayah said:

Tedious was the word i was looking for but I couldnt think of it, I think ill stick to my ram for now and might just buy a better kit at somepoint hopefully ddr4 prices come more down with ddr5 coming hopefully a bit more relevant with the new processors. Thank you for all the help and suggestions, appreciate it a ton!

DDR4 used might come down a bit more, but new I'm pretty sure it's bottomed out, or very close to bottoming out. Still, look on eBay to see if you can find a good kit for cheap. The cheapest kits that are fun to overclock are either Crucial Ballistix, anything 3200MHz CL14, or Corsair memory kits with "v4.31" on the sticker. 

[zayah]

11 minutes ago, RONOTHAN## said:

DDR4 used might come down a bit more, but new I'm pretty sure it's bottomed out, or very close to bottoming out. Still, look on eBay to see if you can find a good kit for cheap. The cheapest kits that are fun to overclock are either Crucial Ballistix, anything 3200MHz CL14, or Corsair memory kits with "v4.31" on the sticker. 

I hear people talk alot about like ''b-die'' what is that and how do I know which kits are that?

[RONOTHAN##]

5 minutes ago, zayah said:

I hear people talk alot about like ''b-die'' what is that and how do I know which kits are that?

Samsung B die is a certain memory IC that can overclock extremely well. It is hilariously inconsistent (I've had kits that can barely hit 3600MHz and I've had kits that have kits that do 5GHz), but as long as you're not buying OEM sticks, I'd be surprised if it didn't overclock at least somewhat decently. 

 

One way to tell is to buy a memory bin that's guaranteed to be Samsung B die. Almost every bin with 3200MHz or higher with flat or mostly flat primary timings (the first 3 primary timings are the same, 14-14-14 for example, or the 2nd two timings are one higher than tCL, 14-15-15 for example). 3200MHz CL14-14-14 is B die, 3600MHz CL14-15-15 is B die, 4400MHz CL19-19-19 is B die, 4000MHz CL18-22-22 is not B die.

 

The other way is to buy a memory maker that lists the memory IC present on the memory chips. The biggest one is Corsair. Corsair has a version number on the sticker that specifies exactly what memory IC it has. There is a way to decode it manually, but the easiest way is to just google it and see what comes up, and the main one to know is v4.31 which is Samsung B die. This trick only works on used memory though, so if you want to buy new you have to find a guaranteed bin of B die. 

[zayah]

12 hours ago, RONOTHAN## said:

Samsung B die is a certain memory IC that can overclock extremely well. It is hilariously inconsistent (I've had kits that can barely hit 3600MHz and I've had kits that have kits that do 5GHz), but as long as you're not buying OEM sticks, I'd be surprised if it didn't overclock at least somewhat decently. 

 

One way to tell is to buy a memory bin that's guaranteed to be Samsung B die. Almost every bin with 3200MHz or higher with flat or mostly flat primary timings (the first 3 primary timings are the same, 14-14-14 for example, or the 2nd two timings are one higher than tCL, 14-15-15 for example). 3200MHz CL14-14-14 is B die, 3600MHz CL14-15-15 is B die, 4400MHz CL19-19-19 is B die, 4000MHz CL18-22-22 is not B die.

 

The other way is to buy a memory maker that lists the memory IC present on the memory chips. The biggest one is Corsair. Corsair has a version number on the sticker that specifies exactly what memory IC it has. There is a way to decode it manually, but the easiest way is to just google it and see what comes up, and the main one to know is v4.31 which is Samsung B die. This trick only works on used memory though, so if you want to buy new you have to find a guaranteed bin of B die. 

Very informative thank you so much!!! Mine is as u descibed 18-22-22-22 so indeed not b-die

[RONOTHAN##]

1 hour ago, zayah said:

Very informative thank you so much!!! Mine is as u descibed 18-22-22-22 so indeed not b-die

There are a lot of memory bins I listed that are B die, so don't just see a kit of something like 3600 CL16-16-16 and think it's not B die because I didn't mention it. The best kit of B die I own, for example, is a kit rated at 3600MHz CL17-18-18, a bin that if gotten used is likely B die, though if bought new today is likely something else. 

[zayah]

3 hours ago, RONOTHAN## said:

There are a lot of memory bins I listed that are B die, so don't just see a kit of something like 3600 CL16-16-16 and think it's not B die because I didn't mention it. The best kit of B die I own, for example, is a kit rated at 3600MHz CL17-18-18, a bin that if gotten used is likely B die, though if bought new today is likely something else. 

I checked a store page of a local pc shop, and they list some of kits in the description if it is a b-die and on some of the kits if it isnt listed in the feedback part the customers say its a b-die kit. 

[RONOTHAN##]

Just now, zayah said:

I checked a store page of a local pc shop, and they list some of kits in the description if it is a b-die and on some of the kits if it isnt listed in the feedback part the customers say its a b-die kit. 

One thing about the non-B die bins is that they can occasionally have B die on them, so just because one person got B die doesn't mean that everyone will get B die. 4000MHz CL18-22-22-42, for example, can occasionally be B die, but most of the time it's not.

[Guest]

7 minutes ago, RONOTHAN## said:

One thing about the non-B die bins is that they can occasionally have B die on them, so just because one person got B die doesn't mean that everyone will get B die. 4000MHz CL18-22-22-42, for example, can occasionally be B die, but most of the time it's not.

Not even occasionally. Never actually.

 

CL 18 bdie will be 18-19-19 4000mhz. I've never seen any other timing sets for bdie at cl18. And this is on par with 4267 and 4400mhz the CL is 19, so 19-19-19, just because it's a little higher frequency and nothing else. Or they'd have to bin the CL18 4400 (as an example) with a higher Vdimm.

[RONOTHAN##]

11 minutes ago, Guest 5150 said:

Not even occasionally. Never actually.

 

CL 18 bdie will be 18-19-19 4000mhz. I've never seen any other timing sets for bdie at cl18. And this is on par with 4267 and 4400mhz the CL is 19, so 19-19-19, just because it's a little higher frequency and nothing else. Or they'd have to bin the CL18 4400 (as an example) with a higher Vdimm.

It can technically happen if they don't have anything else on hand when they go to make the kits, though it is indeed very uncommon. It will never be guaranteed to be B die, though there are kits with B die that are rated at 4000MHz CL18-22-22 or worse

 

https://www.ebay.com/itm/203938995939?hash=item2f7bb61ee3%3Ag%3ADIsAAOSwgVJic5dd&LH_ItemCondition=4&nma=true&si=mWjGhFWi1ThDLQ5ucTpe8Cnx8G8%3D&orig_cvip=true&nordt=true&rt=nc&_trksid=p2047675.l2557

 

This kit, for example, is 4000MHz CL19-23-23, but if you look at the version number (v4.31), it's Samsung B die. 

 

EDIT:

Same with this kit, 3200MHz CL16-20-20, definitely not a B die bin, but it does have B die according to the version number

https://www.ebay.com/itm/394096775039?hash=item5bc1ff8f7f%3Ag%3AXA8AAOSwSqhilQ73&LH_ItemCondition=4&nma=true&si=mWjGhFWi1ThDLQ5ucTpe8Cnx8G8%3D&orig_cvip=true&nordt=true&rt=nc&_trksid=p2047675.l2557

Edited June 14, 2022 by RONOTHAN##

[Guest]

19 minutes ago, RONOTHAN## said:

It can technically happen if they don't have anything else on hand when they go to make the kits, though it is indeed very uncommon. It will never be guaranteed to be B die, though there are kits with B die that are rated at 4000MHz CL18-22-22 or worse

 

https://www.ebay.com/itm/203938995939?hash=item2f7bb61ee3%3Ag%3ADIsAAOSwgVJic5dd&LH_ItemCondition=4&nma=true&si=mWjGhFWi1ThDLQ5ucTpe8Cnx8G8%3D&orig_cvip=true&nordt=true&rt=nc&_trksid=p2047675.l2557

 

This kit, for example, is 4000MHz CL19-23-23, but if you look at the version number (v4.31), it's Samsung B die. 

 

EDIT:

Same with this kit, 3200MHz CL16-20-20, definitely not a B die bin, but it does have B die according to the version number

https://www.ebay.com/itm/394096775039?hash=item5bc1ff8f7f%3Ag%3AXA8AAOSwSqhilQ73&LH_ItemCondition=4&nma=true&si=mWjGhFWi1ThDLQ5ucTpe8Cnx8G8%3D&orig_cvip=true&nordt=true&rt=nc&_trksid=p2047675.l2557

So super low end early bdie. 2016? Earlier?

 

The good stuff never sells that cheap. 

 

Revision number could be accurate, but if that's the only way to know without a confirmation, I'd consider it very old bdie or not at all. 

 

I think the stopped production of bdie in 2019, but don't quote me. It would be nice to know the build date of some of these kits, that would be helpful.

[RONOTHAN##]

1 minute ago, Guest 5150 said:

I think the stopped production of bdie in 2019, but don't quote me

Nope, they stopped producing OEM sticks of the stuff, but the actual IC has never left production. 

 

2 minutes ago, Guest 5150 said:

So super low end early bdie. 2016? Earlier?

I've seen these kits more recently as well, just can't find them in my 5 minutes of searching on eBay. They're not super common, and the B die in them is never particularly great, but it's still B die and still clocks like B die. 

 

My entire point by that message is just because on person had gotten a B die kit doesn't mean that they're all B die, and you really need to judge the memory bin itself in order to tell. Even a memory bin that was guaranteed b die for a while (3600MHz CL17-18-18) isn't guaranteed to be B die anymore, so if you find a review from 2-3 years ago, it'll say it's B die, but if you actually go to buy it, it won't be. 

[Guest]

1 hour ago, RONOTHAN## said:

Nope, they stopped producing OEM sticks of the stuff, but the actual IC has never left production. 

 

I've seen these kits more recently as well, just can't find them in my 5 minutes of searching on eBay. They're not super common, and the B die in them is never particularly great, but it's still B die and still clocks like B die. 

 

My entire point by that message is just because on person had gotten a B die kit doesn't mean that they're all B die, and you really need to judge the memory bin itself in order to tell. Even a memory bin that was guaranteed b die for a while (3600MHz CL17-18-18) isn't guaranteed to be B die anymore, so if you find a review from 2-3 years ago, it'll say it's B die, but if you actually go to buy it, it won't be. 

I'm fairly certain bdie stopped being produced. At least for desktop memory modules, I can't find any information with memory dated 2020 or newer with bdie. I read it at some tech site, not sure which one. One the phone so I'm not gonna look. I'll just take your word on it.

 

I've seen typhoon burner incorrect on a number of occasions. Gotta pull the heat spreaders to see the chips. Kinda like with a lot of video cards to know if it's hynix or samsung.

[RONOTHAN##]

2 minutes ago, Guest 5150 said:

I'm fairly certain bdie stopped being produced. At least for desktop memory modules, I can't find any information with memory dated 2020 or newer with bdie. I read it at some tech site, not sure which one. One the phone so I'm not gonna look. I'll just take your word on it.

 

I've seen typhoon burner incorrect on a number of occasions. Gotta pull the heat spreaders to see the chips. Kinda like with a lot of video cards to know if it's hynix or samsung.

It's just the OEM sticks that no longer are B die, Samsung still sells the actual ICs to the likes G.Skill and Corsair.

 

Even pulling the heat spreaders isn't a guaranteed way to tell. A lot of places like G.Skill, Corsair, Patriot, etc. actually have Samsung/Hynix/Micron/etc. put G.Skill/Corsair/etc.'s logo on their instead of the actual IC manufacturer with no markings referring to which specific IC it is. 

[Guest]

1 minute ago, RONOTHAN## said:

It's just the OEM sticks that no longer are B die, Samsung still sells the actual ICs to the likes G.Skill and Corsair.

 

Even pulling the heat spreaders isn't a guaranteed way to tell. A lot of places like G.Skill, Corsair, Patriot, etc. actually have Samsung/Hynix/Micron/etc. put G.Skill/Corsair/etc.'s logo on their instead of the actual IC manufacturer with no markings referring to which specific IC it is. 

Cool Im gonna give that a good watching. Thank you.

It's messed up. But at least the stronger kits you can tell just by looking at the timing sets, and that's good enough for me really. 

 

I did link AMD random restarts to Hynix memory modules. Particularly the lower end sticks. 3000mhz LPX Vengeance. I have 2 sets. At defaults, rare random restarts. At XMP, often random restarts. The cure, replaced by 3600mhz CL16-16-16-36 B-Dies, never had an issue after that. Tested out of 4 or 5 processors as well. 1200, 1400, 200ge and 220ge. I did use them with 2700x, but horrible performance anyways. And I skipped Micron completely on DDR4 yet. But I'd want the E-dies from them considering the better scaling. Looser timings isn't everything either. Lot of people don't understand the CPU overclock also lowers memory latency. 

[xg32]

On 6/13/2022 at 2:39 PM, zayah said:

.

i'd try for 48 cache and not changing anything else, with ambient high, just keep it there.

[zayah]

4 hours ago, xg32 said:

i'd try for 48 cache and not changing anything else, with ambient high, just keep it there.

thank u for help. i tried to run prime95 with 46cache on 1.28v and it blue screened , i tried 47cache on 1.3v it blue screened. I tried 5.1ghz with same 45 cache on 1.31v and i let it ran an hour without any problems. seems to be rly hard for me to run higher cache without upping voltage alot, compared to just upping core clocks . havent tested yet how much id need to attain that tho

[zayah]

On 6/14/2022 at 1:37 AM, RONOTHAN## said:

5GHz 1.27V is a really solid CPU. 

 

Realistically, you won't notice the difference, but at the same time, I would probably do that. There's not much harm in running a 9900K at 95C in Prime95, so I'd probably aim for 5.1GHz core 4.7GHz cache 1.36-1.4V if you can cool it, just for the benchmarks. 

was able to run 5.1ghz 1.31v but seem to have problems with upping the cache past 45 

[Hairless Monkey Boy]

When you are talking about voltages, are you talking about what you set in BIOS, what you get with vdroop, or the VID?

 

With my 9900K at 5GHz all-core (up to 5.3Ghz single) I have my BIOS voltage set to something like 1.37V (I'd have to check it's been a while) with a relatively low load line calibration (LLC) which allows it to vdroop down to ~1.26V in Prime 95 small FFT. VID is around 1.42V in that case. I run my cache at 4.8Ghz, so my cache appears to be stronger than yours, but I get temps around 80-90C on each core while delidded with liquid metal and a custom loop so your cores seem stronger than mine (although my RAM OC may contribute to extra heat idk).

 

In any case, the performance gains you get will be marginal, so the reason to do this is for fun.

 

I'm on an Asus Maximus XI Hero btw, not too different from your Z390-F