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

[Taraquin]

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.

The stuttering is caused by something. In most cases you wont get stuttering. I have build 4 7800X3D setups and had stuttering on 2 of them. One due to unstable fclk, 1 due to some ram timings I ran too aggressive. No stuttering after resolvibg issues. A few UE-titles will get some stuttering early in game due to shadercompiling. More core will handle that better.

As for efficiency. Run it stock with curve optimizer at the best value you get. It will be very efficient then. Setting pbo to -100 will also improve efficiency greatly as the voltagecurve gets very inefficient slightly above 5GHz. 7800X3D beats 9800X3D on efficiency due to this. Max clocks are 5050 vs 5225. That 175MHz accounts for over 100mv more.

I would say the perfect cpu would be a 9950X3D with 3 cache on both ccds and same clockspeed on ccds. It has better bw than 9800X3D although latency is a bit worse. As for intel a slightly downclocked and undervolted 13900K/14900K is also a viable option if you want both good gaming and productivity performance. Main issue with Intel is the need for fast ram which requires a rather expensive MB to handle it while Amd can get very good performance from a B650 at 120usd.

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.

When the scheduling works 3d cache on one ccd is good. When it dont work it sucks. Windows sucks at prioritizing cores, treads etc. Intel still has issues with this in windows and HT, e-cores etc. This is more a Microsoftissue than a Intel/Amd-issue. Even with software from both Intel and Amd problem still persists in certain apps/games.

 

[JustBenching]

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.

Here is the link, it's from the latest TPU review and includes all the CPUs mentioned.

https://tpucdn.com/review/amd-ryzen-7-9800x3d/images/power-multithread.png

Yes, you are correct that the 9800x 3d draws less power (it's 155 watts at stock btw) than the 285k or the 9950x but it's vastly slower and a lot harder to cool cause that power is concentrated on a lot fewer cores. Power draw itself isn't the issue, the problem in those MT workloads we are talking about is that it draws 150w while not performing to the standards of a 150w 2024 500$ CPU.

That's why I suggested the 9950x3d , it's gonna solve this issue and hopefully I don't think it will cost twice the money, for around 699$ (if that's the price) it's totally worth it over 500$ for the 9800x 3d.

 

[ShrimpBrime]

Here is the link, it's from the latest TPU review and includes all the CPUs mentioned.

https://tpucdn.com/review/amd-ryzen-7-9800x3d/images/power-multithread.png

Yes, you are correct that the 9800x 3d draws less power (it's 155 watts at stock btw) than the 285k or the 9950x but it's vastly slower and a lot harder to cool cause that power is concentrated on a lot fewer cores. Power draw itself isn't the issue, the problem in those MT workloads we are talking about is that it draws 150w while not performing to the standards of a 150w 2024 500$ CPU.

That's why I suggested the 9950x3d , it's gonna solve this issue and hopefully I don't think it will cost twice the money, for around 699$ (if that's the price) it's totally worth it over 500$ for the 9800x 3d.

9800X3D isn't made to be a thread monster, this is perfectly clear.

285K is slower for gaming. And since this is a flagship cpu, any part number below it is also slower for gaming.

Earlier in the thread 9800X3D was tuned 110w at +200 pbo. I don't need to provide a link because your posting in the very same thread, just go back some pages and read.

9950X3D is the Intel killer. Oh wait, Intel lost the efficiency war as far back as 2nd gen Ryzen. AMD Cpus where slower, but more power effecient.

And that's what the person was asking about. A cpu that does NOT utilize a lot of energy to operate, but snappy performance. Perhaps this person would be better suited with something that is not a flagship cpu. Budgets are in people's minds after all.

 

[kDude]

The stuttering is caused by something. In most cases you wont get stuttering. I have build 4 7800X3D setups and had stuttering on 2 of them. One due to unstable fclk, 1 due to some ram timings I ran too aggressive. No stuttering after resolvibg issues. A few UE-titles will get some stuttering early in game due to shadercompiling. More core will handle that better.

As for efficiency. Run it stock with curve optimizer at the best value you get. It will be very efficient then. Setting pbo to -100 will also improve efficiency greatly as the voltagecurve gets very inefficient slightly above 5GHz. 7800X3D beats 9800X3D on efficiency due to this. Max clocks are 5050 vs 5225. That 175MHz accounts for over 100mv more.

I would say the perfect cpu would be a 9950X3D with 3 cache on both ccds and same clockspeed on ccds. It has better bw than 9800X3D although latency is a bit worse. As for intel a slightly downclocked and undervolted 13900K/14900K is also a viable option if you want both good gaming and productivity performance. Main issue with Intel is the need for fast ram which requires a rather expensive MB to handle it while Amd can get very good performance from a B650 at 120usd.


When the scheduling works 3d cache on one ccd is good. When it dont work it sucks. Windows sucks at prioritizing cores, treads etc. Intel still has issues with this in windows and HT, e-cores etc. This is more a Microsoftissue than a Intel/Amd-issue. Even with software from both Intel and Amd problem still persists in certain apps/games.

My plan was more like just to enable XMP and touch nothing else.Hopefully I avoid any memory related stutter shenanigans by letting xmp do the timings for me.I have some 30-38-38-96 6000mhz ram selected in my pcpartpicker list already.

 

[ShrimpBrime]

My plan was more like just to enable XMP and touch nothing else.Hopefully I avoid any memory related stutter shenanigans by letting xmp do the timings for me.I have some 30-38-38-96 6000mhz ram selected in my pcpartpicker list already.

I recommend starting a help me thread of your own and list the system specs makes and models of products. The issue. What you've tried and what you haven't tried. I'm at a loss why you are buying ram with CL30 6000. Is it 32gb or 64gb?? Help us help you!

 

[kDude]

I recommend starting a help me thread of your own and list the system specs makes and models of products. The issue. What you've tried and what you haven't tried. I'm at a loss why you are buying ram with CL30 6000. Is it 32gb or 64gb?? Help us help you!

Build is still in waiting(not yet built).I came to this thread initially to check how the 9800X3D ran since OP and other users here are sharing their results.
OP and another user noted random occasional stutters occuring with their 9800X3D and ram was talked about potentially being the cause.I'm trying to figure out what the cause could be before I actually buy the 9800X3D.(i.e if its actually a ram issue,bios or chipset driver)The 9800X3d is fairly new so it could be anything.
As for why cl 30 6000mhz?(32GB) It was the most suggested set of ram specs for stability/performance on other threads.

 

[Wolverine2349]

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.

What else is a good bet that does not have stability and degradation issues ( or 13th and 14th Gen 13900K/14900K/14700K would be good) or the bad memory latency on non 3D cached AMD chips like the R9 7000/9000 series if you want more than 8 cores and good latency.

Seems like options are pretty bad and be stuck with 8 cores and 9800X3D is onky good option. 9950X3D only 1 3D cached CCD. yikes just like 7950X3D and scheduling issues and locked to 8 cores when gaming even if the game can scale beyond it for secondary less important threads.

And Arrow or should I say Error Lake is beyond awful.

 

[JustBenching]

What else is a good bet that does not have stability and degradation issues ( or 13th and 14th Gen 13900K/14900K/14700K would be good) or the bad memory latency on non 3D cached AMD chips like the R9 7000/9000 series if you want more than 8 cores and good latency.

Seems like options are pretty bad and be stuck with 8 cores and 9800X3D is onky good option. 9950X3D only 1 3D cached CCD. yikes just like 7950X3D and scheduling issues and locked to 8 cores when gaming even if the game can scale beyond it for secondary less important threads.

And Arrow or should I say Error Lake is beyond awful.

If you are manually tuning raptorlake why would it have stability and degradation issues? It doesn't have any.

 

[Wolverine2349]

If you are manually tuning raptorlake why would it have stability and degradation issues? It doesn't have any.

Well whats a safe vcore and current. I have had RPL CPUs that I manually tuned and swapped and returned many. All were manually tuned and temps were in check. They passed stress tests. Some max temps 95C some only in 80s. Some undervolted. They seemed stable then weeks later Cinebench Run or Shader compilation WHEA or BSOD whenb it passed those same tests without issue only a couple weeks prior

I decided to give another go around last March and all was good until Cinebench App error with only a slight undervolt when it passed everything else easily with max 85C temp even toughest Y CRUNCHER SFT and BKT torture was last straw for me.

Did you read and follow the news just a few months back about all the vendors having crashing in servers even when they were at stock settings. And Raptor Lake was internally an Intrel project to be done in case Meteor Lake was not ready and was rushed.

It has issues.

Prove otherwise.

 

[phanbuey]

Actually raptorlake is a steal right now becuase people are scared of it... agree with @JustBenching - if you set static voltage you end up with a great chip for cheap that performs really well (especially on windows 10).

If it was me though i would go for the 7950X used for like $400 with a $100-$120 used/refurb B650 or B650E board and a cheaper 64gb 6000 of hynix-a and just tune it - I would look to then upgrade again on zen 6 to the 10 or 12 core X3d depending on the CCD layout and sell the 7950X.

AMD Ryzen™ 9 7950X 16-Core, 32-Thread Desktop Processor 730143314534 | eBay

 

[Wolverine2349]

Actually raptorlake is a steal right now becuase people are scared of it... agree with @JustBenching - if you set static voltage you end up with a great chip for cheap that performs really well (especially on windows 10).

If it was me though i would go for the 7950X used for like $400 with a $100 used B650 or B650E board and a cheaper 64gb 6000 of hynix-a and just tune it - I would look to then upgrade again on zen 6 to the 10 or 12 core X3d depending on the CCD layout and sell the 7950X.

I heard static vcore is actually worse and ued all my chips tht way and returned and sold off many that developed issues down the road. I read that on overclock.net forums LAST March after trying a few in late 2022 through mid 2023 with DDR4 and DDR5 boards as Ihad a DDR5 stability nightmare issues before swithcing to 7800X3D in SUmmer 2023 when stability issues popped up once more. I decided last March to try with an auto vcore and power limit set on MSI Z690 Unif X (Oh RAM XMP stable with 2 DIMM board that way) and it seemed better until the Cinebench app error and that was last straw for me.

Maybe I could get a 9950X and no hybrid dual CCD as each CCD the same and hope to get 6400 RAM 1:1. Even then it loses to Intel 13th and 14th Gen and blown away by 8 core X3Ds, until games come out that meaningfully scale beyond 8 cores or the few that do. Though 4K maybe not much difference in any game RTX 4090 or 5090 and beyond????

Though 8 cores seems small for RTX 5090 upgrade and beyond n future if I want to keep the same core system (mobo CPU RAM) for 5 years for 4K gaming.

 

[phanbuey]

I heard static vcore is actually worse and ued all my chips tht way and returned and sold off many that developed issues down the road. I read that on overclock.net forums/. I decided last March to try with an auto vcore and power limit set and it seemed better until the Cinebench app error and that was last straw for me.

Maybe I could get a 9950X and no hybrid dual CCD as each CCD the same and hope to get 6400 RAM 1:1. Even then it loses to Intel 13th and 14th Gen and blown away by 8 core X3Ds, untiol games come out that meaningfully scale beyond 8 cores or the few that do. Though 4K maybe npot much difference.

Though 8 cores seems small for RTX 5090 upgrade and beyond n future if I want to keep the same core system (mobo CPU RAM) for 5 years for 4K gaming.

Yeah it's true that's why ended up going for the 9800X3D ultimately. It's a decent mix of everything, and should last a good while with the 4090, I plan on skipping 5 series altogether since the 4090 is more than enough for current gen, and I'm not above medium settings.

I did get this stable to 2200 IF and 6400 1:1 cl30 which is smooth as butter, and I also accidentally managed for one boot get my effective clocks to 5.7Ghz stable to run through a bunch of Corona runs (didn't get a chance to stress test before losing the tuning), it really seems like there's more performance left in this chip once the bioses stabilize.

 

[rv8000]

I heard static vcore is actually worse and ued all my chips tht way and returned and sold off many that developed issues down the road. I read that on overclock.net forums/. I decided last March to try with an auto vcore and power limit set on MSI Z690 Unif X (Oh RAM XMP stable with 2 DIMM board that way) and it seemed better until the Cinebench app error and that was last straw for me.

Maybe I could get a 9950X and no hybrid dual CCD as each CCD the same and hope to get 6400 RAM 1:1. Even then it loses to Intel 13th and 14th Gen and blown away by 8 core X3Ds, until games come out that meaningfully scale beyond 8 cores or the few that do. Though 4K maybe not much difference in any game RTX 4090 or 5090 and beyond????

Though 8 cores seems small for RTX 5090 upgrade and beyond n future if I want to keep the same core system (mobo CPU RAM) for 5 years for 4K gaming.

There’s almost no difference at 4k, investing in a dead platform like raptor lake and older is foolish.

You can’t really “future-proof” anything aside from buying a socket thats going to have drop in replacements; jury is out on z890, AM5 is at least guaranteed Zen 6 at this point.

8 cores will remain relevant until next console generation at a minimum. If all your doing is playing some games + disc + maybe streaming most high core count chips are relatively pointless.

 

[JustBenching]

Well whats a safe vcore and current. I have had RPL CPUs that I manually tuned and swapped and returned many. All were manually tuned and temps were in check. They passed stress tests. Some max temps 95C some only in 80s. Some undervolted. They seemed stable then weeks later Cinebench Run or Shader compilation WHEA or BSOD whenb it passed those same tests without issue only a couple weeks prior

I decided to give another go around last March and all was good until Cinebench App error with only a slight undervolt when it passed everything else easily with max 85C temp even toughest Y CRUNCHER SFT and BKT torture was last straw for me.

Did you read and follow the news just a few months back about all the vendors having crashing in servers even when they were at stock settings. And Raptor Lake was internally an Intrel project to be done in case Meteor Lake was not ready and was rushed.

It has issues.

Prove otherwise.

I had the chip, what more do I need to prove? With a unify X that you mentioned.

 

[Wolverine2349]

I had the chip, what more do I need to prove? With a unify X that you mentioned.

You had a 13900K? Was it platinumn stable? I imagine Unify X, you got XP stable just fine as it is 2 DIMMER. But was other things platinum like Cinebench Y CRUNCHER shader compilation? Stable for a while no degradation? Did you use contact frame. What kind of cooler?

Is it possible contact frame causes instability due to mounting pressure?

I use a Noctua NH-D15. Can mounting pressure form large dual tower coolers cause issues I wonder?

Someone mentioned before 6-8 months ago (Do not remember where) that they had Cinebench app error and tried a differnet mobo and fixed it and wodnered if maybe mobo or random thing was cause. That was after I threw in towel when things seemed to be going well last March until that upon my final tuning! just to see Cinebnech score aftr pasisng much tougher stress tests.

 

[Taraquin]

Intel has admitted some batches og 13th gen had oxidation issues, that is not something that can be fixed by manually downvolting/underclocking. There is probably a bit of individual variance in what is safe voltages as temperature is factor that can contribute so at very low temps they may be more tolerant to higher voltages. If you buy a 14th gen now and use latest microcode etc it should be 100% safe.

 

[Wolverine2349]

Intel has admitted some batches og 13th gen had oxidation issues, that is not something that can be fixed by manually downvolting/underclocking. There is probably a bit of individual variance in what is safe voltages as temperature is factor that can contribute so at very low temps they may be more tolerant to higher voltages. If you buy a 14th gen now and use latest microcode etc it should be 100% safe.

Are todays 14th Gen chips produced better and with microcode update 100% safe. Were batches of 14th Gen produced a few months ago or longer possibly not safe. Did Intel fix manufacture ring issues with oxidation issues.

How many of the chips actually had the oxidation issues?

 

[Taraquin]

Are todays 14th Gen chips produced better and with microcode update 100% safe. Were batches of 14th Gen produced a few months ago or longer possibly not safe. Did Intel fix manufacture ring issues with oxidation issues.

How many of the chips actually had the oxidation issues?

With latest bios you get it atleast. Oxidation was fixes some unknown date in 2023. How many cpus affected Intel has not disclosed.

 

[JustBenching]

Been playing POE2 and sadly performance is pretty horrible. Drops to below 60 in some heavier fights

 

[Taraquin]

Been playing POE2 and sadly performance is pretty horrible. Drops to below 60 in some heavier fights

Due to cpu or gpu? I havent tried it yet, but I expect a rough ride the first few weeks atleast.

 

[JustBenching]

Due to cpu or gpu? I havent tried it yet, but I expect a rough ride the first few weeks atleast.

CPU, 1 core hits 99% usage. I've seen drops to below 40 too

Here is a short video I made

 

[Taraquin]

CPU, 1 core hits 99% usage. I've seen drops to below 40 too

Here is a short video I made

Poor distribution of load it seems. Only 1 fully loaded and 2 partially. Rest sleeps. Nice consumption atleast

 

[kDude]

CPU, 1 core hits 99% usage. I've seen drops to below 40 too

Here is a short video I made

That's gotta be the game right?

 

[JustBenching]

That's gotta be the game right?

Hope so

 

[RootinTootinPootin]

Are todays 14th Gen chips produced better and with microcode update 100% safe. Were batches of 14th Gen produced a few months ago or longer possibly not safe. Did Intel fix manufacture ring issues with oxidation issues.

it was only a specific batch, pretty sure the 2024 batches are safe even the early 2022 13th gen chips.