• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Ryzen 7 9800X3D

  • Thread starter Thread starter Hostile_18
  • Start date Start date
Different take using DLSS which gives interesting results.

This sort of stuff is exactly why there is
A) No one true king for gaming right now, as whilst the 9800X3D is faster overall, usually in averages, and sometimes its quite dominant in lows, and requires less tuning (and cheaper RAM) to get the best out of it (a lot less), it does still have weaknesses where Raptor Lake does better.
B) It is still worth tweaking RAM timings, at least primaries and secondaries, for gaming, even with an X3D chip, as it still comes into play whenever the cache cannot be leveraged properly; and the high averages of the X3D make the minimums when it falls outside it's comfort zone all the more notable; and in those worst case scenarios, it can actually have a notable impact.

Honestly; I don't like Framechasers' attitude, and he is wrong sometimes, at the same time, a lot of people just disregard him completely because of his attitude, but he's not always wrong, and he's been commenting on this for a while. Do I think he always gets it right, no, but I suspect we will see more of this sort of commentary over time, like this video from DannyZ, as people start to see there are more fringe cases and edge scenarios than there used to be, because CPUs have become a lot more complicated.

Even he does not disagree the 9800X3D is a lot more power efficient in many cases, and having now got both in the same household, I do understand his criticism of the 7800X3D; its a baller gaming chip, but its notably slower around windows than a tuned modern Intel, or 9*** series chips; and can have heavier dips, especially if the RAM is not tweaked to reduce latency; and when latency is tuned on both, even using Buildzoid's easy timings, the 9000 and Intel DO feel that bit snappier; due to a combo of lower system latency, or in Intels' case, monolithic, but even the 9800X3D can suffer dips when the vCache is not able to be monopolised on, because realistically there is just more latency in a chiplet architecture than a monolithic one without VERY clever and advanced engineering. And if rumours are true, AMD have paid heed to that and are going to be reducing physical distance between chips, and improving the interconnect (IF/Sea of Wires) and IO Die itself heavily in 10000 series; whether those rumours come true is another thing though, as I don't like peddling rumours as they're so often wrong, but for gaming especially, and general latency, this is an incredibly low hanging fruit, so it'd be logical it might be true.

Intel 2** series however suffers all the issues AMD had with Chiplets, and they've got less experience tuning them. If I was buying for gaming now, I'd 100% still go a 9000 series AMD, or a 14*** series Intel.

I'd argue, especially out of the box, with the power on top, the 9800X3D is the right chip for most people, price being right, as you need to tune a lot more with Intel, BUT there is still a place for Intel, even more so if you do productivity rather than just gaming on the system; and if it was a choice between a 9600X and a 14600K, unless I was 100% serious about upgrading down the road, I'd probably actually lean towards the 14600K; now the BIOS updates have come out to stop them killing themselves with overaggressive single core boost and voltages.
 
Last edited:
@LtMatt Not really debunked then, more a case, its a fringe situation; there ARE still situations where either power monitoring, latency etc can cause worse frame dips with AMD than Intel. (can cause issues with Intel too on some platforms, the joy of so many different platform variables!!)

As above, I still think the AMD chip is the superior choice for most; but its not QUITE as bulletproof in these fringe scenarios and interactions as Intel traditionally has been, especially with the recommended config.

You are also running MUCH faster memory clocks on AMD than most; and are way outside the oft recommended 6000C30; along with a 2200IF, which some chips (mine included) won't even boot at using relatively safe voltages, which may also be helping you here.
I saw a noticeable difference going from the standard recommended config, to my tuned config, and you're considerably further ahead than that again; so I'm not sure your configuration is really that representative of the average X3D user's experience. (plus you're running a 9950X3D not a 9800X3D, so different RAM bandwidth too, beyond your lower realworld latency from the 8200C34)

It's good to know your experience is much better than the examples given, again highlighting the memory/IO subsystem is the likely bottleneck which you're managing to negate which that very nice tuned RAM and IF, but you're also running a considerably faster system than the examples in some ways; and in some ways proving there are limitations, that I'd like to hope the 10000 series AMD chips will address!
 
Last edited:
I said apparently as reflex works fine for me, so it’s 100% debunked IMO. Others have tested it too on OCN and also got similar results to me, so it’s not just some magical tuning that I pulled out of my behind.

I was running 6400Mhz in that video I posted earlier so nothing too special, either. And FClK makes barely any difference these days. The results speak for themselves though.

60% better 0.1% lows using the same cpu in my video. Insert Harold gif. :p
 
Last edited:
Heh, he's just uploaded a new video and has fixed his lows. Seems like it was related to monitoring a specific sensor in MSI AB all along. :cry:
 
I said apparently as reflex works fine for me, so it’s 100% debunked IMO. Others have tested it too on OCN and also got similar results to me, so it’s not just some magical tuning that I pulled out of my behind.

I was running 6400Mhz in that video I posted earlier so nothing too special, either. And FClK makes barely any difference these days. The results speak for themselves though.

60% better 0.1% lows using the same cpu in my video. Insert Harold gif. :p

Ah, no worries, I noticed in the description of your video it still mentioned 8200c34 so didnt realise you'd knocked that on its head, fair point in that regards.

(The video says this:
"I waited until all 91 bots spawned into the map for maximum CPU stress.

User.Cfg File Settings
(no image quality tweaks)
Thread.ProcessorCount16Thread.MaxProcessorCount 16

9950X3D + 2200Mhz FCLK + PBO
Cooled by Valkyrie SYN 360 AIO + 100% Fan Speed + Liquid Metal
Asus X670E Gene 3304 BIOS
DDR5 8200Mhz C34
5090 Aorus Xtreme AIO + 220Mhz Core +3000Mhz Memory
Nvidia 581.57 1080P Low Preset
Windows 11 25H2
LG C4 OLED 48"
Recorded Via Shadow Play (recording cost is 5-+ FPS) .")

The guy has posted a followup around 45 minutes ago mentioning he's got to the bottom of the issue, still it does seem like some AMD configs do take more of a hit due to sensors, as he highlights Unwinder also gave feedback previously, so there is still a question as to why those sensor pollings have a larger detriment than they do on Intel. Not sure whether you've got any perspective there?

He is right on that a LOT of gamers have overlays on these days from one software or another, so the issue there is still relevant.

Edit: Just saw your follow up. The joys of starting writing a post a while before you actually post it lol
 
Last edited:
Ah, no worries, I noticed in the description of your video it still mentioned 8200c34 so didnt realise you'd knocked that on its head, fair point in that regards.

(The video says this:
"I waited until all 91 bots spawned into the map for maximum CPU stress.

User.Cfg File Settings
(no image quality tweaks)
Thread.ProcessorCount16Thread.MaxProcessorCount 16

9950X3D + 2200Mhz FCLK + PBO
Cooled by Valkyrie SYN 360 AIO + 100% Fan Speed + Liquid Metal
Asus X670E Gene 3304 BIOS
DDR5 8200Mhz C34
5090 Aorus Xtreme AIO + 220Mhz Core +3000Mhz Memory
Nvidia 581.57 1080P Low Preset
Windows 11 25H2
LG C4 OLED 48"
Recorded Via Shadow Play (recording cost is 5-+ FPS) .")

The guy has posted a followup around 45 minutes ago mentioning he's got to the bottom of the issue, still it does seem like some AMD configs do take more of a hit due to sensors, as he highlights Unwinder also gave feedback previously, so there is still a question as to why those sensor pollings have a larger detriment than they do on Intel. Not sure whether you've got any perspective there?

He is right on that a LOT of gamers have overlays on these days from one software or another, so the issue there is still relevant.

Edit: Just saw your follow up. The joys of starting writing a post a while before you actually post it lol
I’ve recorded so many videos of that cpu benchmark looks like I shared the wrong one. Here's the one i meant to share running 6400Mhz and only the cache CCD so as to mimic a 9800X3D.

Yeah not sure about that sensor, I don’t monitor that sensor on MSI AB as it’s well documented to cause an issue on the 9800X3D. I monitor the CPU power sensor via HWINfO64 which has no such issue. What’s most odd is DannyZ uploaded a video around a year ago highlighting that same issue, so why he created this video in the first place is a bit odd when he already knew about the power sensors on MSI AB lol. Anyway alls well that ends well I guess. :p
 
Last edited:
ok then I mean that one, so why aren't they used as 16 core CPU's?
They are, however when gaming you don't want the game running on the non cache ccd, one because it doesn't have the vcache and also because the intercore latency can cause stutter and drops in fps.

I switched from a 5900x which had two ccds and it was a great chip however I didn't know how smooth gaming could be on pc, for the two reasons stated above.

It's also the reason I went with the 9800x3d and won't upgrade until they release a 12 core ccd vcache. I know that the scheduling and game bar stuff do work to keep the game on one ccd well these days but I don't need 16 cores yet and we'll wait to see how it all pans out.
 
Last edited:
They are, however when gaming you don't want the game running on the non cache ccd, one because it doesn't have the vcache and also because the intercore latency can cause stutter and drops in fps.

I switched from a 5900x which had two ccds and it was a great chip however I didn't know how smooth gaming could be on pc, for the two reasons stated above.

It's also the reason I went with the 9800x3d and won't upgrade until they release a 12 core ccd vcache. I know that the scheduling and game bar stuff do work to keep the game on one ccd well these days but I don't need 16 cores yet and we'll wait to see how it all pans out.
5900x user here and im planning on going am5 real soon using the 9800x3d as a stop gap until the rumoured 12 core single ccd x3d zen 6 part if the rumour is true happens
so looks like im on a similar trajectory just a step and a bit behind
 
5900x user here and im planning on going am5 real soon using the 9800x3d as a stop gap until the rumoured 12 core single ccd x3d zen 6 part if the rumour is true happens
so looks like im on a similar trajectory just a step and a bit behind
It's a very worthy upgrade. I was worried about losing 12 cores and was going to jump on a 9900X, but @Tetras advised the 9800x3d and all I can say is don't worry about it. It's a very fast cpu and we can both be happy when the 12 core ccd comes out in a few years.
 
yeah i doubt id get away with the am4 x3d part as i do sometimes use more than just hte 8cores when doing larger music projects but i think with the higher individual core capabilities on the zen5 x3d chips i think id be fine on the 9800x3d and well when zen 6 comes around..... fingers crosed il find some excuse as to why i need to go back to 12 cores but without the dual ccd drawbacks
 
yeah i doubt id get away with the am4 x3d part as i do sometimes use more than just hte 8cores when doing larger music projects but i think with the higher individual core capabilities on the zen5 x3d chips i think id be fine on the 9800x3d and well when zen 6 comes around..... fingers crosed il find some excuse as to why i need to go back to 12 cores but without the dual ccd drawbacks

probably wait until see what's up I have 7700 8 core so I'd want more cores, X3D, single CCD.
 
oh i deffo want the 12 core on one ccd and the v cache lol but in the meantime im still on am4 so i need to get a am5 rig up and running in the nexct couple of months and i guess il be using the 9800x3d until the fabled hopefully actually real 12 core thingy comes out
as i dont think the 2x6 layout of the 9900x3d is worth the tradeoff and my 5900x although great just doesnt have those extra capabilities and its a bit old tech now lol
and for the first time in my life i actually want something a bit more closer to vaguely up to date for once
tbh im shocked im not still using a cpc 464 amstrad or something lol
 
oh i deffo want the 12 core on one ccd and the v cache lol but in the meantime im still on am4 so i need to get a am5 rig up and running in the nexct couple of months and i guess il be using the 9800x3d until the fabled hopefully actually real 12 core thingy comes out
as i dont think the 2x6 layout of the 9900x3d is worth the tradeoff and my 5900x although great just doesnt have those extra capabilities and its a bit old tech now lol
and for the first time in my life i actually want something a bit more closer to vaguely up to date for once
tbh im shocked im not still using a cpc 464 amstrad or something lol
Yeah all the niceties of AM5 are worth it. More nvme drives and faster memory to begin with. Make sure you pick a good board with little lane sharing if you can.
 
Back
Top Bottom