14900k - Tuned for efficiency - Gaming power draw

[Panther_Seraphin]

Could they also be doing this because HT has been vulnerable to attacks in the past? Is this still a viable question with HT on, or has this problem been fixed up via CPU micro codes?

This is the real reason.

HT in the enterprise environment has proved time and again in recent times to be one of if not the easiest attack vectors currently known especially with it being used to bypass virtualisation security by being able to access whatever is running on the other thread on the same CPU when its not from the same VM.

AMD have already got offerings in via Zen 4c (EPYC 9754S) that suits these use cases to please major cloud providers/heavy virtualisation users.

XMP is another bucket of worms. If the compatibility is poor the user experience may be riddled with problems (hello AM4, not sure about modern Intel).
Another consumer unfriendly area, although it improved, but still unfriendly.

XMP is an Intel standard that AMD "Supports" but I doubt they have put the a lot of resource into XMP as they have developing their own standard EXPO. Also with memory speeds and timings getting faster and tighter the differences in memory controller performance and optimization highlights how a "one size fits all" approach to memory isnt working as it once was.

 

[BoggledBeagle]

This is the real reason.

HT in the enterprise environment has proved time and again in recent times to be one of if not the easiest attack vectors currently known especially with it being used to bypass virtualisation security by being able to access whatever is running on the other thread on the same CPU when its not from the same VM.

A good thing about 14900K is that it has so many e-cores, that it really does not need HT and if HT will need to be disabled in the future, I will not notice it at all.

 

[dgianstefani]

A good thing about 14900K is that it has so many e-cores, that it really does not need HT and if HT will need to be disabled in the future, I will not notice it at all.

That and the security reasons. Future chips will have 32 etc E cores.

Mobile chips already have three groups of cores, the P cores, the E cores, and the two E cores on a low power island with the IO so the rest of the chip can be power gated. This is coming to desktop too.

 

[fevgatos]

As an update, TLOU again after applying some undervolting, power draw is around 90-95w. Keep in mind this is by far the heaviest game in terms of power draw, running completely CPU limited. In other games or more normal gaming scenarios the power draw is 50-60 watts.

14900k - 5.5 ghz HT OFF

Just for comparisons sake, a 7800x 3d and a 12900k below. The 7800x 3d consumed the least amount of power (75-80 watts) but it was the slowest of the 3. The 14900k consumed more power than the 3d but less than the 12900k but it was decisively beating both in performance. All rigs are memory tuned with 4090 gpus.

Spoiler

The 14900k shows incredible advantages in power draw over the 12900k due to mostly binning. It can hit those insane clocks with minimum voltages. Realistically, I could further tune the 12900k to drop power to the same 80-90 watts but it's already losing in performance vs the 14900k, making it more efficient would just make it as slow as the 7800x 3d. Now I'd like to get a 14900ks and test if there is any increase there too in terms of binning

 

[Taraquin]

As an update, TLOU again after applying some undervolting, power draw is around 90-95w. Keep in mind this is by far the heaviest game in terms of power draw, running completely CPU limited. In other games or more normal gaming scenarios the power draw is 50-60 watts.

14900k - 5.5 ghz HT OFF

Just for comparisons sake, a 7800x 3d and a 12900k below. The 7800x 3d consumed the least amount of power (75-80 watts) but it was the slowest of the 3. The 14900k consumed more power than the 3d but less than the 12900k but it was decisively beating both in performance. All rigs are memory tuned with 4090 gpus.

Spoiler

The 14900k shows incredible advantages in power draw over the 12900k due to mostly binning. It can hit those insane clocks with minimum voltages. Realistically, I could further tune the 12900k to drop power to the same 80-90 watts but it's already losing in performance vs the 14900k, making it more efficient would just make it as slow as the 7800x 3d. Now I'd like to get a 14900ks and test if there is any increase there too in terms of binning

Was the 7800X3D also undervolted? You can shave off atleast 10-15W with a -20 to -40 CO which all samples I have tried can do easily. Good thing for the future is that Intel ditches HT on next gen. For many productivity cases it has its uses, but in gaming it seems like there generally is more trouble than gain.

 

[fevgatos]

Was the 7800X3D also undervolted? You can shave off atleast 10-15W with a -20 to -40 CO which all samples I have tried can do easily. Good thing for the future is that Intel ditches HT on next gen. For many productivity cases it has its uses, but in gaming it seems like there generally is more trouble than gain.

It was, but TLOU is excessively heavy in power draw. It's by far the most intense game that's why im testing this for my undervolts / power draws - cause in any other game the power draw is between 40 and 80 watts even on the 14900k.

Good thing for the future is that Intel ditches HT on next gen. For many productivity cases it has its uses, but in gaming it seems like there generally is more trouble than gain.

Yeah, for gaming HT on Intel is useless / damaging, it already has Ecores when it needs the extra oomph. Especially the power draw increase is huge with no actual performance increase in games.

 

[londiste]

Was the 7800X3D also undervolted? You can shave off atleast 10-15W with a -20 to -40 CO which all samples I have tried can do easily. Good thing for the future is that Intel ditches HT on next gen. For many productivity cases it has its uses, but in gaming it seems like there generally is more trouble than gain.

I have a 7800X3D sample that started causing issues at -15 CO...

 

[fevgatos]

I have a 7800X3D sample that started causing issues at -15 CO...

Yeah it's silicon lottery, but realistically it wouldn't make much difference either way. It would drop power from 80 to what, 75 watts?

 

[Taraquin]

I have a 7800X3D sample that started causing issues at -15 CO...

Maybe 1 core sucked?

Yeah it's silicon lottery, but realistically it wouldn't make much difference either way. It would drop power from 80 to what, 75 watts?

one sample did avg CO if -41 (worst core -32). That is about 15-20W in the games I tested. My current do -36 avg(worst core -22), last sample I tested did -30(worst core -25) avg.

It was, but TLOU is excessively heavy in power draw. It's by far the most intense game that's why im testing this for my undervolts / power draws - cause in any other game the power draw is between 40 and 80 watts even on the 14900k.


Yeah, for gaming HT on Intel is useless / damaging, it already has Ecores when it needs the extra oomph. Especially the power draw increase is huge with no actual performance increase in games.

Helldivers 2 is also quite cpu-heavy, but reproducing results are difgicult

 

[fevgatos]

Maybe 1 core sucked?

one sample did avg CO if -41 (worst core -32). That is about 15-20W in the games I tested. My current do -36 avg(worst core -22), last sample I tested did -30(worst core -25) avg.

15-20w from stock to - 41 right? But what about - 20 to - 40, the difference will be probably be tiny - irrelevant. I'm doing - 20 on my 6900hs and the difference in gaming is not really noticeable.

 

[Taraquin]

15-20w from stock to - 41 right? But what about - 20 to - 40, the difference will be probably be tiny - irrelevant. I'm doing - 20 on my 6900hs and the difference in gaming is not really noticeable.

Yeah, a good CO can make quite a bit of difference to consumption\temps. -20 is more in the 5-10W territory and not that important I agree. On notebooks lower noise and better GPU boost due to chared cooling is more important

 

[fevgatos]

Yeah, a good CO can make quite a bit of difference to consumption\temps. -20 is more in the 5-10W territory and not that important I agree. On notebooks lower noise and better GPU boost due to chared cooling is more important

Ι always get the banana chips, so my lap crashes at idle with - 20

Still the problem here with the x3d isn't the power, it's drawing the least, the problem is the performance.

 

[Taraquin]

Ι always get the banana chips, so my lap crashes at idle with - 20

Still the problem here with the x3d isn't the power, it's drawing the least, the problem is the performance.

Yeah, in some titles the cache aint king or the lack of cores is a bigger issue. i recommend you try HD2 for testing cpu performance. My friends supertuned 14900K (ht off) with 8400 ram is about 5% faster, but uses over 150W, my 7800X3D uses 55-65W.

 

[fevgatos]

Yeah, in some titles the cache aint king or the lack of cores is a bigger issue. i recommend you try HD2 for testing cpu performance. My friends supertuned 14900K (ht off) with 8400 ram is about 5% faster, but uses over 150W, my 7800X3D uses 55-65W.

Dont have the game but I might get it just to test. How would I actually test a multi game though?

 

[the54thvoid]

Just to remind everyone, this thread is about tuning the 14900k for better gaming efficiency (against itself). It's not about AMD.

Let's keep it on track, and keep it helpful.

(I've removed a post because it was pitching red versus blue.)

 

[Taraquin]

Dont have the game but I might get it just to test. How would I actually test a multi game though?

Testing is diffucult atm, not build in bench, but you see on utilization that it uses all you bring to the table. It is also a really fun game

 

[fevgatos]

Deleted, sorry.

Testing is diffucult atm, not build in bench, but you see on utilization that it uses all you bring to the table. It is also a really fun game

None of my friends play it so I haven't tried it myself, do you have a video of it running on your system so I can replicate it? Preferably at the start of the game

 

[Taraquin]

Deleted, sorry.


None of my friends play it so I haven't tried it myself, do you have a video of it running on your system so I can replicate it? Preferably at the start of the game

No, I dont :/ Start of the game is tutorial, not demanding at all. Demanding is when 60 aliens soawn at once while you drop 3 bombardments and using sentries

 

[fevgatos]

No, I dont :/ Start of the game is tutorial, not demanding at all. Demanding is when 60 aliens soawn at once while you drop 3 bombardments and using sentries

Game is almost downloaded. Was looking online and it doesn't seem like the 3d cache makes a lot of difference to this game. Is this video valid or it's complete BS?

Spoiler

What settings am I supposed to use? All ultra? Im thinking about opening a new thread for it but dunno in what section

 

[iameatingjam]

I haven't had much trouble getting my new 14700kf to where I want it to be. Mind you, I only play my games at 60hz, and thats not necessarily a big ask for a chip like this. I used to undervolt too, but since my first chip died after one month of undervolting.... It could be a coincidence, but I'd rather not take the chance again. Besides, limiting my chips to just 253w, while asking 5.5ghz all core in a gaming workload, seems rather safe in my experience. Haven't had issues in the last 6 months. The only difference being... the undervolt. I know it goes against pretty much all common knowledge. So it could be coincidence, or it could just be luck of the draw. Who knows?

 

[Taraquin]

Game is almost downloaded. Was looking online and it doesn't seem like the 3d cache makes a lot of difference to this game. Is this video valid or it's complete BS?

Spoiler

What settings am I supposed to use? All ultra? Im thinking about opening a new thread for it but dunno in what section

I use a combo of high/ultra. Try res scaling and watch the 'magic' hapoend shout out if you wanna do a few rounds yeah that scaling is valid from what I have seen. Both my 7800X3D and my friends 14900K get hammered with res scaling, still getting good fps though, 160-280 on my setup, my friend about 10fps more.

 

[fevgatos]

I use a combo of high/ultra. Try res scaling and watch the 'magic' hapoend shout out if you wanna do a few rounds yeah that scaling is valid from what I have seen. Both my 7800X3D and my friends 14900K get hammered with res scaling, still getting good fps though, 160-280 on my setup, my friend about 10fps more.

Tried it yesterday, everything ultra with low resolution just to push the cpu, it's not that bad, people were going crazy about how heavy it is. Dropped to 150 when i was in a 4 man team, around 180-200 when solo, that's on a stock 12900k.I expected drops below 100 to be fair

Spoiler

I haven't had much trouble getting my new 14700kf to where I want it to be. Mind you, I only play my games at 60hz, and thats not necessarily a big ask for a chip like this. I used to undervolt too, but since my first chip died after once month of undervolting that could be a coincidence, but I'd rather not take the chance again. Besides, limiting my chips to just 253w, while asking 5.5ghz all core in a gaming workload, seems rather safe in my experience. Haven't had issues in the last 6 months. The only difference being... the undervolt. I know it goes against pretty much all common knowledge. So it could be coincidence, or it could just be luck of the draw. Who knows?

UV can theoretically (very theoretically) kill your mobo - maybe - in extreme scenarios, but probably it will never happen. But killing your CPU i'm almost certain it can't happen.

 

[Taraquin]

Tried it yesterday, everything ultra with low resolution just to push the cpu, it's not that bad, people were going crazy about how heavy it is. Dropped to 150 when i was in a 4 man team, around 180-200 when solo, that's on a stock 12900k.I expected drops below 100 to be fair

Spoiler

UV can theoretically (very theoretically) kill your mobo - maybe - in extreme scenarios, but probably it will never happen. But killing your CPU i'm almost certain it can't happen.

When you get to 9higher difficulties in 4 player and you have 50+ enemies + massive bombardment, that is when cpu load starts to climb. It was at hard-missions and beyond my 7800X3D vent above 70C using 65W while gaming. Solo/duo and eady to medium difficulty is not taxing

Watched a few videos on your channel, good work man Interesting to see TLOU utilizing ecores quite well it seems.

 

[fevgatos]

When you get to 9higher difficulties in 4 player and you have 50+ enemies + massive bombardment, that is when cpu load starts to climb. It was at hard-missions and beyond my 7800X3D vent above 70C using 65W while gaming. Solo/duo and eady to medium difficulty is not taxing

If you can share some actual footage from you or your friend so I can do my tests over the weekend would be great. I didn't play much yesterday so have no idea how to get to higher difficulties and all that.

 

[Taraquin]

If you can share some actual footage from you or your friend so I can do my tests over the weekend would be great. I didn't play much yesterday so have no idea how to get to higher difficulties and all that.

I can try You need to level up etc to unlock harder missions. May be a few days before I can. What do you record with? Team up on a match or 2 and we can compare real time in same match if you want?