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

[JustBenching]

x3D cache adds ~5ns to the AIDA result.

The main nvme has a direct connection to the CPU if I'm not mistaken, the second nvme goes through the chipset if the SSD is SATA, probably also through the chipset.

The fine-tuned 12900k has more cores (threads) and more raw performance, so in heavy scenes with many objects - it will be faster.
The 9800x3D will be faster where latency plays a role because in most cases it will use the cached x3D data (~7ns) but not the RAM (~60ns).

12900k isn't faster, it's just gets less of a hit in performance in heavier scenes. It's still slower than the 9800x 3d (bar the odd game here and there) but as I've said, way more consistent. I think it's totally due to the memory subsystem. A cache hit imposes a way bigger penalty on the x3d due to the slower memory subsystem, but a correct cache prediction also helps it more, resulting in insanely high maximum frames but relatively low minimum frames.

View attachment 373277

new AGESA is out - got a bit of latency reduction.

View attachment 373276

Frame pacing is a bit better. Still beta bios tho.

You are getting 64 after the new agesa or before? Cause im stuck at 63-64 with 6400c28 tuned to hell. AMD promised an agesa fix for that.

 

[AVATARAT]

12900k isn't faster, it's just gets less of a hit in performance in heavier scenes. It's still slower than the 9800x 3d (bar the odd game here and there) but as I've said, way more consistent. I think it's totally due to the memory subsystem. A cache hit imposes a way bigger penalty on the x3d due to the slower memory subsystem, but a correct cache prediction also helps it more, resulting in insanely high maximum frames but relatively low minimum frames.

Yes, it's situational so it will only happen in some scenes, the average will be lower for 12900k.

*

You are getting 64 after the new agesa or before? Cause im stuck at 63-64 with 6400c28 tuned to hell. AMD promised an agesa fix for that.

On my 9700x latency is up ~10ns too, and yeah we wait for patch. That's happen after AGESA 1.2.0.2a.

 

[Taraquin]

1.2.0.2a introduced some latency issue afaik, upwards of 7ns. Still present in 1.2.0.2b.

6400 c28-30 with relatively tight timings should be around 58-59ns or better. While the 7900X3D and 9800X3D are different chips, some of my better latency results hit the high 57s at 6400 c30 2133 fclk on the 7900X3D.

Stutter in games can also be a result of fclk instability.

On my 7800X3D I get consistent 1ns better latency with 1.2.0.2a vs previous agesas

 

[phanbuey]

This is 1.2.0.2b.

on 1.2.0.2a my min latency was 68.5ns this dropped it to 64

 

[phanbuey]

New Bios - last max was 440 fps wish substantially lower mins.

Effective clocks are reading at 5.7 Ghz....

Scores are substantially higher... PBO off - A-XMP on but primary timgns set to auto. Chip is also running cooler -- doesnt break 78C. Stutters are now virtually gone but board suuuper picky with ram settings, faster is almost never better when it comes to primary timings, letting it auto set yields the highest scores. Effective clock is reporting everywhere (hwinfo, benchmarks) at 5.7ghz and actual clock is clocking in at 5.25....

 

[JustBenching]

Im too scared to update the bios, it will most likely delete my saved profile and then Ill have to go through memory tuning again - and most likely ill need to tinker with some timings etc. I might get it over with during christmas.

Now having the chip for a couple of weeks I feel like the point I was making the last several years has been on point. amd's chips aren't particularly efficient, they are just limited to very low power limits out of the box. When you are trying to match the power draw they are substantially less efficient than most intel chips. This chip is substantially slower in every MT workload at the same power as my 12900k (besides ycruncher, it's faster there due to AVX). The 12900k is a 2021 chip...

 

[AVATARAT]

Im too scared to update the bios, it will most likely delete my saved profile and then Ill have to go through memory tuning again - and most likely ill need to tinker with some timings etc. I might get it over with during christmas.

You can take ZenTimings and print/image RAM timings with your phone and then set it in the BIOS (AMD).

 

[JustBenching]

You can take ZenTimings and print/image RAM timings with your phone and then set it in the BIOS (AMD).

Yeah I know that, I just don't think the same timings will train with the new bios and same nitro settings. . Don't have to take a screenshot, I remember them all

 

[Taraquin]

View attachment 373959
New Bios - last max was 440 fps wish substantially lower mins.

Effective clocks are reading at 5.7 Ghz....
View attachment 373960

Scores are substantially higher... PBO off - A-XMP on but primary timgns set to auto. Chip is also running cooler -- doesnt break 78C. Stutters are now virtually gone but board suuuper picky with ram settings, faster is almost never better when it comes to primary timings, letting it auto set yields the highest scores. Effective clock is reporting everywhere (hwinfo, benchmarks) at 5.7ghz and actual clock is clocking in at 5.25....

As for ram rfc and refi are the 2 big ones really worth tuning

Im too scared to update the bios, it will most likely delete my saved profile and then Ill have to go through memory tuning again - and most likely ill need to tinker with some timings etc. I might get it over with during christmas.

Now having the chip for a couple of weeks I feel like the point I was making the last several years has been on point. amd's chips aren't particularly efficient, they are just limited to very low power limits out of the box. When you are trying to match the power draw they are substantially less efficient than most intel chips. This chip is substantially slower in every MT workload at the same power as my 12900k (besides ycruncher, it's faster there due to AVX). The 12900k is a 2021 chip...

For MT 12900K will be king for many years vs a 8c/16t chip I think viritually everyone buys X3D-cpus for gaming though. Amd tuned the clock/voltagecurve and ppt quite efficient. Intel was not bad at 12th gen, but 13/14th is just bad out of the box unless you use ppt and reminds me of Amd vs Nvidia with Vega 64 vs gtx 1080.

Much of the efficiencytuning with Amd can bw done by adjusting soc, vddp etc voltages. Going from 1.1v soc to 1.3v soc can increase consumption at full load by over 20W tanking efficiency.

 

[JustBenching]

As for ram rfc and refi are the 2 big ones really worth tuning


For MT 12900K will be king for many years vs a 8c/16t chip I think viritually everyone buys X3D-cpus for gaming though. Amd tuned the clock/voltagecurve and ppt quite efficient. Intel was not bad at 12th gen, but 13/14th is just bad out of the box unless you use ppt and reminds me of Amd vs Nvidia with Vega 64 vs gtx 1080.

Much of the efficiencytuning with Amd can bw done by adjusting soc, vddp etc voltages. Going from 1.1v soc to 1.3v soc can increase consumption at full load by over 20W tanking efficiency.

It's not just multi though, having steam download a game + watching a video on YouTube the x3d is casually hitting 60 watts. The 900k was at 1/3 of that. A lot of it is indeed the soc voltage but even with the stock 1.1 it's still hitting 40s. Even sitting there idle with 1.25v soc I'm at 35 watts!! The 900k was at low single digits.

 

[Taraquin]

It's not just multi though, having steam download a game + watching a video on YouTube the x3d is casually hitting 60 watts. The 900k was at 1/3 of that. A lot of it is indeed the soc voltage but even with the stock 1.1 it's still hitting 40s. Even sitting there idle with 1.25v soc I'm at 35 watts!! The 900k was at low single digits.

As for idle-consumption I agree as for what is reported, but 35W at idle sounds very high, what powerplan? Lots of background resources? My 7800X3D sits at 18W idle at 1.09v soc. At 1.27v soc it idled at 24W. In cinebench however the difference was above 20W.

If I enable igpu consumption at idle drops to 13W due to lower voltages\clocks at idle and wallconsumption drops to 49W.

My 12400F used around 5W idle. This is much due to the IO-die+ccd-design which makes very low idlepower impossible. When I measured from the wall however both setups idles at about 56W w\o monitor. Maybe more actions are handled via MB on the Intel-part and not cpu package?

I noticed the same with my 8400 and my 5600X. Lower reported packageconsumption reported in windows (5 vs 16W), but wallplug reported about identical draw. In all cases similar psu (650W gold and 2dimm matx MB was used along with 1 nvme-drive and 2 casefans.

At very low ppt Intel will excel due to monodesign, but I wonder how actual powerdraw from wall is since I had 2 intelsystems where MB apparently used more than Amd-counterparts.

 

[JustBenching]

As for idle-consumption I agree as for what is reported, but 35W at idle sounds very high, what powerplan? Lots of background resources? My 7800X3D sits at 18W idle at 1.09v soc. At 1.27v soc it idled at 24W. In cinebench however the difference was above 20W.

If I enable igpu consumption at idle drops to 13W due to lower voltages\clocks at idle and wallconsumption drops to 49W.

My 12400F used around 5W idle. This is much due to the IO-die+ccd-design which makes very low idlepower impossible. When I measured from the wall however both setups idles at about 56W w\o monitor. Maybe more actions are handled via MB on the Intel-part and not cpu package?

I noticed the same with my 8400 and my 5600X. Lower reported packageconsumption reported in windows (5 vs 16W), but wallplug reported about identical draw. In all cases similar psu (650W gold and 2dimm matx MB was used along with 1 nvme-drive and 2 casefans.

At very low ppt Intel will excel due to monodesign, but I wonder how actual powerdraw from wall is since I had 2 intelsystems where MB apparently used more than Amd-counterparts.

Always balanced power plan, no background processes at all, I'm using custom lean windows installs. Only hwinfo is running. I can tell by the temps alone that the power draw is accurate, cpu idles at 15c+ higher than my intel did. With a room temp of 21C the intel idles at 30 to 35c, the 9800x 3d is at 47 to 51.

 

[rv8000]

Always balanced power plan, no background processes at all, I'm using custom lean windows installs. Only hwinfo is running. I can tell by the temps alone that the power draw is accurate, cpu idles at 15c+ higher than my intel did. With a room temp of 21C the intel idles at 30 to 35c, the 9800x 3d is at 47 to 51.

Something is very wrong with your setup.

19c ambient, nh-dh15 one fan at 1500rpm, pbo max and positive offsets with 105 bclk & 1.25 vsoc my 9800x3D idles at 30-31c. Your idle temps are higher than some of my gaming load temps.

 

[JustBenching]

Something is very wrong with your setup.

19c ambient, nh-dh15 one fan at 1500rpm, pbo max and positive offsets with 105 bclk & 1.25 vsoc my 9800x3D idles at 30-31c. Your idle temps are higher than some of my gaming load temps.

You have the fan running at 1500 rpm on idle? Mine is at 0.

I don't know, maybe my cooler doesn't like amd.

 

[Taraquin]

Always balanced power plan, no background processes at all, I'm using custom lean windows installs. Only hwinfo is running. I can tell by the temps alone that the power draw is accurate, cpu idles at 15c+ higher than my intel did. With a room temp of 21C the intel idles at 30 to 35c, the 9800x 3d is at 47 to 51.

That is unusually high. My 7800X3D idles at around 35C with 19C ambient and fans running at very low speed.

 

[rv8000]

You have the fan running at 1500 rpm on idle? Mine is at 0.

I don't know, maybe my cooler doesn't like amd.

Set that way for ram overclock stability testing atm. Even at 500 rpm it’s ~32c.

I guess using a subpar hsf and running a fan stop profile for idle you’re going to have horrible temps, but why even do that? Most fans are silent around 800 or less rpm.

 

[JustBenching]

Set that way for ram overclock stability testing atm. Even at 500 rpm it’s ~32c.

I guess using a subpar hsf and running a fan stop profile for idle you’re going to have horrible temps, but why even do that? Most fans are silent around 800 or less rpm.

I'm doing it because temps were perfectly fine on my 12900k - 13900k and 14900k with the same subpar HSF and fan stop.

Now I paid some more attention in hwinfo, the CPU CCD temperatures are fine, it idles between 27 and 32c (0 rpm fans), it's the io die that's extremely high. Right now with nothing open and the fan at 750rpm it's at 42-43C.

At 1.27v soc it idled at 24W.

At 1.25vsoc mine sits at 24-26w, so similar numbers, I just realized having steam open on the 9800x 3d just increases power draw by 10-15w , lol.


Just did some testing, maxed out fans at idle, the core temperature drops to 25c, the l3 is at 26, but IOD stays at 39-40c.

 

[phanbuey]

You have the fan running at 1500 rpm on idle? Mine is at 0.

I don't know, maybe my cooler doesn't like amd.

Yah on idle my fans are like 600 rpm...

My idle temps are the same as yours, takes a bit to cool down after gaming session. With the new bios update it locks clocks at 5.37Ghz (testing enhanced PBO boost w/o voltage bump) and rarely breaks 75C -- fan rpm to 1200RPM at load.

 

[Taraquin]

I'm doing it because temps were perfectly fine on my 12900k - 13900k and 14900k with the same subpar HSF and fan stop.

Now I paid some more attention in hwinfo, the CPU CCD temperatures are fine, it idles between 27 and 32c (0 rpm fans), it's the io die that's extremely high. Right now with nothing open and the fan at 750rpm it's at 42-43C.


At 1.25vsoc mine sits at 24-26w, so similar numbers, I just realized having steam open on the 9800x 3d just increases power draw by 10-15w , lol.


Just did some testing, maxed out fans at idle, the core temperature drops to 25c, the l3 is at 26, but IOD stays at 39-40c.

There is a bug with steam causing constant load on cpu. I noticed mine idling around 25-30W with steam open. Latest steam released fixed it for me. Idles around 18W again. Epic has never had any issue with idleload.

 

[kDude]

Are you guy with the 9800x3d's experiencing occasional system stutters?

 

[JustBenching]

Are you guy with the 9800x3d's experiencing occasional system stutters?

In gaming? Yeah, but it's very rare, happens once every 20 minutes + in some games.

 

[phanbuey]

Are you guy with the 9800x3d's experiencing occasional system stutters?

Not system stuttering anymore - but I was getting these on the initial beta bios - mouse would freeze for one second or so.

 

[kDude]

In gaming? Yeah, but it's very rare, happens once every 20 minutes + in some games.

Both gaming and outside gaming.

Not system stuttering anymore - but I was getting these on the initial beta bios - mouse would freeze for one second or so.

But you don't get them anymore I assume since you said initial beta bios?

I'm asking because I noticed a bunch of threads on reddit saying the same thing with their 9800X3D's.System stutters and in games.Wondering if its OS related or bios.Cpu is still new so maybe it needs more updates.

 

[JustBenching]

Both gaming and outside gaming.

Outside gaming haven't experienced any slowdowns or anything of the likes that some people "cough cough" are reporting (the amd lag etc.). Works fine, there is a small hit on how fast it launches some apps but I assume that's mostly because I downgraded my OS Nvme from a 980pro to something cheap I have no idea what it is and what it's called. Still, not anything noticeable, it's very snappy and so was my 12900k, no complaints.

 

[phanbuey]

Both gaming and outside gaming.

I mean if it happens randomly id consider it as stuttering.But you don't get them anymore I assume since you said initial beta bios.

I'm asking because I noticed a bunch of threads on reddit saying the same thing.System stutters/etc...Could be w11/multiple monitors causing it or bios needing more cooking time.

yeah i had these initially but havent had them since updating bios/chipsets and windows tweaks - not sure what stopped it but definitely don't have them anymore.