9800x 3d vs 12900k - Battle of the Century

[kDude]

Can you also try disabling above 4G decoding and Resizeable Bar? For science.

Unoptimized direct access I/O to video card can cause some gnarly stutters.

Are these 2 settings enabled by default in recent new motherboards?I have a 5 year old mobo and that stuff is off by default.
Edit:My mobo doesn't even have gamebar setting .4g decoding is there tho but disabled by default
Also disabling ftpm seems to be another possible "fix" for the stutter.I still keep seeing threads as fresh as 1 month.How is this still an issue.

 

[mb194dc]

Calling it hot and slow while using PBO and overclocking, then comparing it to a heavily undervolted 12900k with more cores and threads isn’t what i’d call a reasonable conclusion. But then again, it wouldn’t be a fevgatos thread without the intel bias behind the curtain.

9800x3d is great for some ultra niche use cases. For the same money you can get 8 extra cores on AM5, or save and get a better gpu instead instead though.

Reviews with a 4090 and benching at 1080p or less / upscaled equivalent, pretty dumb to me since doesn't match 99%+ of real world use cases.

Op just telling it how it is and looking past the over hyping.

 

[phanbuey]

Are these 2 settings enabled by default in recent new motherboards?I have a 5 year old mobo and that stuff is off by default.
Edit:My mobo doesn't even have gamebar setting .4g decoding is there tho but disabled by default
Also disabling ftpm seems to be another possible "fix" for the stutter.I still keep seeing threads as fresh as 1 month.How is this still an issue.

They're enabled by default on B650 boards and X670, not sure about the brand new ones but I would assume yes. FTPM doesn't really cause stutters on AM5 -- i bought a TPM module just in case, but it makes no difference for me if im using external module or fTPM, it was a huge issue on AM4 iirc.

 

[mkppo]

M-die is less stressful on the imc, better chance for 8000mhz if your chip is good and you’ve got a proper board.

Ultimately a-die is faster. Tighter trfc, rrds and rrdl, and can clock well, but is more likely to require more vdd/vddq.

Dual rank sticks also offer a slight advantage in certain games/applications, but will require additional tertiary timing tuning.

I'm just done testing on 6200 GDM D and going to shoot for 8000 soon on a HDV with 24GB M-die kit. Any tips on how much looser the RRDS/L are compared to A die? I know about TRFC but just wondering what other timings I should look out for

 

[AVATARAT]

Ill test for a bit today with 1800 to see if it's any more frequent and than the 2133, although it doesn't seem to be the case.

As I said above, in most cases stutters can be bypassed with a RAM tuning, in some specific situations, with few steps lower than 2200 fabric.

 

[kDude]

They're enabled by default on B650 boards and X670, not sure about the brand new ones but I would assume yes. FTPM doesn't really cause stutters on AM5 -- i bought a TPM module just in case, but it makes no difference for me if im using external module or fTPM, it was a huge issue on AM4 iirc.

You sure am5 are not effected?I've seen some users on reddit amdhelp with their 7800X3D resolving their stuttering issues by disabling ftpm.Honestly he might aswell try disabling it and see how it goes.At best frametimes are better and at worst nothing changes.

 

[rv8000]

I'm just done testing on 6200 GDM D and going to shoot for 8000 soon on a HDV with 24GB M-die kit. Any tips on how much looser the RRDS/L are compared to A die? I know about TRFC but just wondering what other timings I should look out for

8 for rrds and 12 for rrdl seem to be reliable minimums. Twtrs seems to bottom out at 6 for both my m-die kits.

Only old reference I have on my phone close to stability in 2:1 with m-die, with a shit board. Youll probably have much better luck with the hdv.

Real VDD/VDDQ from ZT = (v listed-0.8)/2 + 0.8

You may also require juiced vddio 1.35-1.45

 

[JustBenching]

View attachment 371555

Shadow of the benchmarks run with some slowish ram at 6000 30-38-38-96

I seem to be getting better scores at 6000 cl30 than 6400 at the same timings....

So 6000/2000/3000 > 6400/2133/3200 .... not sure why - both sets stable.

Im getting better numbers with FCLK @ 1800 then with FCLK @ 2133.

CPU average is 472 with 1800, 469 @ 2133. But at least it's stutter free in both cases

As I said above, in most cases stutters can be bypassed with a RAM tuning, in some specific situations, with few steps lower than 2200 fabric.

You sure am5 are not effected?I've seen some users on reddit amdhelp with their 7800X3D resolving their stuttering issues by disabling ftpm.Honestly he might aswell try disabling it and see how it goes.At best frametimes are better and at worst nothing changes.

Im running 6400c28 tuned and FTPM is off. Anyways, don't really care that much at the moment, I don't think im keeping the chip so whatever, let it stutter

9800x3d is great for some ultra niche use cases. For the same money you can get 8 extra cores on AM5, or save and get a better gpu instead instead though.

Reviews with a 4090 and benching at 1080p or less / upscaled equivalent, pretty dumb to me since doesn't match 99%+ of real world use cases.

Op just telling it how it is and looking past the over hyping.

Thanks man, I really don't know what Im supposed to say to not be biased. Like I've tested the chips, in MT the x3d is a lot slower while pulling the same power, how am I supposed to convey that without sounding biased?

I can post some results if someone is interested , tried corona, cinebenches, indigo, povray and I don't know what else. The x3d needed around 50% more power to match the 12900k (150 vs 100 watts), it was only faster in ycruncher with avx enabled.

 

[Taraquin]

After having it for a couple of weeks

Gaming, it's fast. Both out of the box and after tuning, it beat my tuned 12900k in everything bar one game. If kept at stock power draw during games is acceptable, can definitely hit over 110w in some games but in general it's pretty well behaved. With +200mhz from the pbo menu power draw and temps go through the roof, I do not recommend, it's pulling more power than my 12900k and it's much harder to cool. Also you only get 4% performance when fully cpu bound-for twice the power, nah. Just leave it at stock. I've noticed some repeated stuttering in games that have no stutters, I'll test some more with stock fclk cause at this point it starts to bother me. Especially when I'm playing dota 2, stuttering at key moment ain't good. It's very rare, happens once every 15 - 20 minutes or more but it's still bothering me.

For normal working (a lot of browsers, excels, pdfs etc) it's as snappy as my 12900k, no issues. I have seen complaints about amd chips taking long to launch apps etc., haven't seen anything like that.

For some heavier multitasking it's kind of a mixed bag. It's not horrible but for the price, oh well, everything else is a better option. The problem here is that even the power draw is humongous, especially for the performance offered. At stock (unless my mobo overclocks the chip at stock, have no idea) it can easily pull 150+ watts, at which point it just thermal throttles at 95c. Reminds me of Intel chips but in a worse way, at least those were pulling a lot of power but were fast, this one pulls power but it's slow. Definitely recxomend the 105 eco enabled from the bios, or a 120ppt limit.

Is it worth the money? If you are playing MSFS, absolutely yes. If you are not, not really.

Eg1. Very important point I forgot, at least with my combination of mobo + cpu, stable settings are... well, always stable. If you are into ram overclocking you know what I mean. You can play around with any settings you want and then just load your stable saved profile, it will boot and train instantly and still remain stable. That's contrary to some raptorlake + apex combinations I had

Good reflexions! As for stuttering I had the same with my 7800X3D with previous ram config. It was caused by errorcorrection due to bad timings that appeared stable in testing. It can be caused by fclk, ram timings, compability issues with some hw/sw etc. Too low soc/vddp-voltage can cause it etc. Generally stuttering is some sort of errorcorrection in my experience. I experience no stuttering now except a few games right after launch due to shadwr compiling etc, but that is solved after a few seconds.

Im getting better numbers with FCLK @ 1800 then with FCLK @ 2133.

CPU average is 472 with 1800, 469 @ 2133. But at least it's stutter free in both cases



Im running 6400c28 tuned and FTPM is off. Anyways, don't really care that much at the moment, I don't think im keeping the chip so whatever, let it stutter


Thanks man, I really don't know what Im supposed to say to not be biased. Like I've tested the chips, in MT the x3d is a lot slower while pulling the same power, how am I supposed to convey that without sounding biased?

I can post some results if someone is interested , tried corona, cinebenches, indigo, povray and I don't know what else. The x3d needed around 50% more power to match the 12900k (150 vs 100 watts), it was only faster in ycruncher with avx enabled.

I think you have issues with 6400 and soc tolerance. Try getting a tight 6200 stable. It will allow you to run faster fclk due to lower soc reqs.

 

[kDude]

Im getting better numbers with FCLK @ 1800 then with FCLK @ 2133.

CPU average is 472 with 1800, 469 @ 2133. But at least it's stutter free in both cases

So you managed to fix the stutter?Did reducing the flck help or was it something else?

Good reflexions! As for stuttering I had the same with my 7800X3D with previous ram config. It was caused by errorcorrection due to bad timings that appeared stable in testing. It can be caused by fclk, ram timings, compability issues with some hw/sw etc. Too low soc/vddp-voltage can cause it etc. Generally stuttering is some sort of errorcorrection in my experience. I experience no stuttering now except a few games right after launch due to shadwr compiling etc, but that is solved after a few seconds.

Where you manually inputting your timings when you ran into the problem?

 

[Taraquin]

So you managed to fix the stutter?Did reducing the flck help or was it something else?

Where you manually inputting your timings when you ran into the problem?

Yes.

 

[kDude]

Yes.

Hopefully I don't run into the problem if I just use expo.Memory related stutters seem like a huge hassle to figure out

 

[JustBenching]

So you managed to fix the stutter?Did reducing the flck help or was it something else?

No, still stutters with 1800. It's hard to test though cause they aren't very frequent to begin with.

 

[kDude]

No, still stutters with 1800. It's hard to test though cause they aren't very frequent to begin with.

Well shit.
Seeing similar occasional stutters on youtube benchmarks displaying a frametime graph aswell.Think il hold out on buying the 9800X3D till it matures a bit more.The random stutters have to be chipset/bios related at this point.
Not like I could buy one if I wanted anyway They're out of stock/overcharging everywhere.

 

[Womble4lard]

If you are talking about gaming efficiency, yea at stock it should be very high up. If you are talking about Mt efficiency, nah, not really. Only if you operate every other chip at insanely high power limits is that the case. Running cbr23 the 12900k is faster while pulling less power

Yes, first slot on most boards, next to the CPU is direct to CPU. Lower slots are chipset.

12900K slower than 14700K

9800X3D dude tuned to 110w +200 Boost.
VS
14100F 4.5ghz 4c8t rated 110w.

12900K smashed. 253w
285K double smashed 250w 8 P-threads
14900K smashed, just way high wattage.
(edit: forgot to mention, doesn't matter what the ram is doing)
10s of FPS differences using RTX 4090 monster! Just Intel costs more to operate.
Drowning in wattage usage during in an age of "going green battery operated lawn equipment".

Buying from scratch, not wanting a energy bill monster, a solid all rounder PC, likely with a 4080 OR its equivalent from 5000. Wanting snappy experience with background tasks running... Is a new 12th gen actually the best bet?

 

[JustBenching]

Buying from scratch, not wanting a energy bill monster, a solid all rounder PC, likely with a 4080 OR its equivalent from 5000. Wanting snappy experience with background tasks running... Is a new 12th gen actually the best bet?

What's your budget? I think the 9950x 3d will be the goat, but if that's too expensive I'd say a 265k or 245k would also be good alternatives. I love my 12th gen, I think it's amazing, but it can't compete with newer cpus. Unless you can find it for cheap, it ain't worth it.

 

[phanbuey]

Buying from scratch, not wanting a energy bill monster, a solid all rounder PC, likely with a 4080 OR its equivalent from 5000. Wanting snappy experience with background tasks running... Is a new 12th gen actually the best bet?

no. 12th gen isn't the best bet unless you want to save money and don't plan on ever upgrading that 4080. It will be an OK rig for the next few years, but you're better off going with a 265K or something similar than 12th gen, it's EOL at this point and not really worth building around unless you're getting it for very little $.

I think the 9950x3d is going to be like the 7950x3d now that it seems like they are not doing cache on both tiles.

 

[ShrimpBrime]

Buying from scratch, not wanting a energy bill monster, a solid all rounder PC, likely with a 4080 OR its equivalent from 5000. Wanting snappy experience with background tasks running... Is a new 12th gen actually the best bet?

12900K is a 253w chip, this is not a consideration today.

AMD 9950x monster at 170w or 9800X3D at 120w. Seems to be limited supply, so the pricing is still high.

Intel is hard to sell. 285K is 250w and only 8 performance core threads. Sure, area effecient cores are much better, but no suitable for gaming in my opinion, and the pre-release reviews and after release reviews show that.

Could always look at 7600X3D chip and upgrade on AM5 later. This platform will extend into 2026.

 

[JustBenching]

12900K is a 253w chip, this is not a consideration today.

AMD 9950x monster at 170w or 9800X3D at 120w. Seems to be limited supply, so the pricing is still high.

Intel is hard to sell. 285K is 250w and only 8 performance core threads. Sure, area effecient cores are much better, but no suitable for gaming in my opinion, and the pre-release reviews and after release reviews show that.

Could always look at 7600X3D chip and upgrade on AM5 later. This platform will extend into 2026.

Those are TDP numbers you are quoting. 9950x draws 220w, 9800x 3d draws 150w. 285k is 235w. TDP isn't powerdraw.

The above numbers are from TPUs review.

 

[Womble4lard]

I appreciate the advice.
The budget isn't too restrictive, it just changes my expectation to extend the life of the system the more I put in. I was thinking maybe a 12900k, let the dust settle for a year or two and hope there's a more obvious choice by then. It surprises me that you both said Ultra's because my very novice research only shows these being hated on and not recommend, Jayztwocents, Gaming nexus, Linus Tech Tips, the ultra's don't get a chance in their recommendations. 265k would be fine budget wise if it's actually a ok buy. Or maybe a 245 to get onto the 1851 socket drop something better into it down the line.... I'm very conflicted which way to go.

I bought a sandybridge 2600k 13 years ago and it was a easy choice to make and has been such a solid work horse... I've been trying to retire it for years but nothing has stood out as the obvious answer.

 

[ShrimpBrime]

Those are TDP numbers you are quoting. 9950x draws 220w, 9800x 3d draws 150w. 285k is 235w. TDP isn't powerdraw.

The above numbers are from TPUs review.

It is exactly the same number.
Thermal design point is the same number of wattage consumed because a processor simply converts electrical energy to thermal energy.

Can say a cpu is 100% effecient converting electrical energy to thermal energy.

9800X3D rated 120w

https://www.amd.com/en/products/processors/desktops/ryzen/9000-series/amd-ryzen-7-9800x3d.html

 

[JustBenching]

It is exactly the same number.
Thermal design point is the same number of wattage consumed because a processor simply converts electrical energy to thermal energy.

Can say a cpu is 100% effecient converting electrical energy to thermal energy.

9800X3D rated 120w

https://www.amd.com/en/products/processors/desktops/ryzen/9000-series/amd-ryzen-7-9800x3d.html

No it's not, AMD measures TDP differently cause it assumes some heat escapes through the sides and not on the side of the cooler. There was a huge article about it but can't be bothered to find it. I can guarantee you, the 9800x 3d does not consume 120w but 150+.

 

[phanbuey]

In eco mode you can lock it to consume 120 or 100W without really losing any gaming performance - but you will lose raw MT performance for when you want to play some cinebench.

The 12900K, especially with an OC, can easily consume 258W+ in the cases where the 9800x3d is eating 150W. My 13700K, at stock, was a 280W chip before it overpowered my AIO and throttled - undervolted I could usually keep it below 230W at full tilt @ stock (5.3ghz all core). But it also was 40% faster in MT so...

in full tilt MT w/perf the 13700K was actually the about the same in MT depending on app. But then again you're not getting this chip for that -- if you want that get the 9950x and undervolt that with a PBO offset for monster workstation efficiency.

 

[TheDigitalJedi]

no. 12th gen isn't the best bet unless you want to save money and don't plan on ever upgrading that 4080. It will be an OK rig for the next few years, but you're better off going with a 265K or something similar than 12th gen, it's EOL at this point and not really worth building around unless you're getting it for very little $.

I think the 9950x3d is going to be like the 7950x3d now that it seems like they are not doing cache on both tiles.

Hearing about this was a downer. Some of my bros at 3D reported that AMD didn't find any significant performance increases with cache on both tiles. I'm still excited for the 9950X3D and curious to see if it will show nice increases in production work over the 7950X.

 

[ShrimpBrime]

No it's not, AMD measures TDP differently cause it assumes some heat escapes through the sides and not on the side of the cooler. There was a huge article about it but can't be bothered to find it. I can guarantee you, the 9800x 3d does not consume 120w but 150+.

OK, say 150w electrical consumption. Still dissipated that figure.

And, still 100w less than any (i7/i9) Intel chip of the last 4 generations now.

And the person asking the question would like a suitable cpu and was the main focus of my comment, but I appreciate your input with or without a link. Thank you.

In eco mode you can lock it to consume 120 or 100W without really losing any gaming performance - but you will lose raw MT performance for when you want to play some cinebench.

The 12900K, especially with an OC, can easily consume 258W+ in the cases where the 9800x3d is eating 150W. My 13700K, at stock, was a 280W chip before it overpowered my AIO and throttled - undervolted I could usually keep it below 230W at full tilt @ stock (5.3ghz all core). But it also was 40% faster in MT so...

in full tilt MT w/perf the 13700K was actually the about the same in MT depending on app. But then again you're not getting this chip for that -- if you want that get the 9950x and undervolt that with a PBO offset for monster workstation efficiency.

Have had my 14700K P cores ONLY at 380w.
Oops typo. 280w my bad.