WWYD? 9900x vs 9700x

Joined
Aug 28, 2021
Messages
940
Im looking at two local cpus...a 9900x that is a lot closer that they're asking $300 and then a 9700x that is $250 but is 60min further away.

My main limit is super modded minecraft with PT shaders...it doesn't matter the gpu, but my 5950x cant keep up...it cant even get 30fps. It is mostly single threaded, but not sure about the two ccds vs just one.

Any thoughts?
 
Your time and gas is bound to be worth more than $50. Get the 9900x. That said, if you're that heavily CPU bound, I would hold out for a 9800X3D, or at least a 7500X3D.

As for single vs dual CCDs, that's more of a problem on the X3D chips than the normal ones.
 
Your time and gas is bound to be worth more than $50. Get the 9900x. That said, if you're that heavily CPU bound, I would hold out for a 9800X3D, or at least a 7500X3D.

As for single vs dual CCDs, that's more of a problem on the X3D chips than the normal ones.
They all do the same thing, try to keep games on the best CCD.

So, a 9900X is mostly a 9600X (with a bit higher clocks), in games. But, a 9600X isn't much worse than a 9700X. So, its not a big deal.
 
Last edited:
Your time and gas is bound to be worth more than $50. Get the 9900x. That said, if you're that heavily CPU bound, I would hold out for a 9800X3D, or at least a 7500X3D.

As for single vs dual CCDs, that's more of a problem on the X3D chips than the normal ones.
That is where I'm at...

Minecraft is all in Java and the consensus is it doesn't utilize extra cache....just pure grunt.
 
They all do the same thing, try to keep games on the best CCD.

So, a 9900X is mostly a 9600X (with a bit higher clocks), in games. But, a 9600X isn't much worse than a 9700X. So, its not a big deal.
Sure, but the Windows scheduler is less likely to jump across CCDs when X3D is not involved and the penalty for jumping isn't as high as on the X3D chips.
That is where I'm at...

Minecraft is all in Java and the consensus is it doesn't utilize extra cache....just pure grunt.
The quick searches I've found show that the extra cache can help in some cases. Given that the 9xxxX3D parts are clocked about the same as their X counterparts, they don't have the performance penalty associated with the 7xxx and 5xxx generations while still having the extra cache to boost performance where it helps. Holding out for a 9800X3D would be the ideal way to go, but also most likely much more expensive than either option you have listed- used market would likely be in the $400 range, though it might drop in the near future with the release of the 9850X3D and 9950X3D2.
 
I haven't seen a game stretch across both CCDs for well over a year now.
 
Always buy the most expensive CPU for your budget. If it is a gaming build, X3D.
 
Jump to 9900x, Minecraft uses a lot of small objects so you need more threads to process them.
Yeah, X3D could do the job too, but I'm not sure which is better in this situation.
 
Sure, but the Windows scheduler is less likely to jump across CCDs when X3D is not involved
For gaming, the goal is the same: Park the less relevant CCD as much as possible. Only wake it up if other things start happening (user opens Firefox to watch youtube on second monitor). For multicore scenarios where the cores are absolutely beneficial (pre-compiling shaders), or wake it up for the few games where crossing to the other CCD is still a net gain for that game.



and the penalty for jumping isn't as high as on the X3D chips.

*Its not that the penalty is higher. Its just that the X3D cache is so much more efficient for most games. So not only do you get the latency hit from crossing to another CCD. But, its alsoo a standard caching setup where more data is now sent out to system RAM; which feels relatively worse.
 
Well, I felt inspired today to investigate how minecraft java handles vcache so I gathered up bits I have in different systems and a (mostly) fresh install of win11.

Board is an aorus x570S with 32GB of ram at 3600MT/s and a 4090. Minecraft is 1.20.1 with Forge and create 5 and running valkyrien skies. For the resource pack, it is hardtop vanilla accurate as well as pixxl. For the shader i used IterationRP 0.8.9. The actual serving of the game is a separate server we use at home.

I ran the same course along the deck of the battleship I am trying to build. It is an assembled physics object to add to the strain. I set the time to 600, then start walking at A turret all long until I'm past Y turret. That carries me over the machinery compartment with the create boilers, steam engines and generators which is where the frames drop off.

Test #1 was a 5950x (this is my good one, not the flaky one) running without PBO.
Textures at 128x 42FPS with 1% lows at 22
Textures at 256x 45FPS with 1% at 22
Textures at 512x 38FPS with 1% at 21

Test #2 was the 5950x except this time running with PBO on at each individual core with a CO value. I think it is a pretty well optimized 5950x
Textures at 128x 42FPS with 1% lows at 20
Textures at 256x 43FPS with 1% at 23
Textures at 512x 38FPS with 1% at 20

The 5950x was never able to push the GPU past 50%.

Test #3 was a 5800x3d running stock curve offset.
Textures at 128x 53FPS with 1% lows at 24
Textures at 256x 48FPS with 1% at 21
Textures at 512x 50FPS with 1% at 23

I was shocked that the optimized 5950x faired no better. I was shocked that the 5800x3d was a lot better. Lastly, I find it interesting that really, the 1% lows don't change at all. All my 5800x3d experience for this was in my main system with a probably seven year install of windows 10 and all the pollution it has acquired where it was worse than the tuned 5950x.
 
If just for gaming, you could probably get enough for even less than both. However, if you do "more" (like homelabbing especially lots of VMs), then more cores wins.
 
If just for gaming, you could probably get enough for even less than both. However, if you do "more" (like homelabbing especially lots of VMs), then more cores wins.
Just gaming and regular use. No heavy cad type stuff.

All my server stuff is on dedicated equipment in the shed.
 
Well, for gaming a 9900X will disable the second CCD turning it into a 6 core CPU (for most game, not all). If you don't have a use case for 12 cores, then I'd recommend going with the 8 core 9700X.
 
Well, I felt inspired today to investigate how minecraft java handles vcache so I gathered up bits I have in different systems and a (mostly) fresh install of win11.

Board is an aorus x570S with 32GB of ram at 3600MT/s and a 4090. Minecraft is 1.20.1 with Forge and create 5 and running valkyrien skies. For the resource pack, it is hardtop vanilla accurate as well as pixxl. For the shader i used IterationRP 0.8.9. The actual serving of the game is a separate server we use at home.

I ran the same course along the deck of the battleship I am trying to build. It is an assembled physics object to add to the strain. I set the time to 600, then start walking at A turret all long until I'm past Y turret. That carries me over the machinery compartment with the create boilers, steam engines and generators which is where the frames drop off.

Test #1 was a 5950x (this is my good one, not the flaky one) running without PBO.
Textures at 128x 42FPS with 1% lows at 22
Textures at 256x 45FPS with 1% at 22
Textures at 512x 38FPS with 1% at 21

Test #2 was the 5950x except this time running with PBO on at each individual core with a CO value. I think it is a pretty well optimized 5950x
Textures at 128x 42FPS with 1% lows at 20
Textures at 256x 43FPS with 1% at 23
Textures at 512x 38FPS with 1% at 20

The 5950x was never able to push the GPU past 50%.

Test #3 was a 5800x3d running stock curve offset.
Textures at 128x 53FPS with 1% lows at 24
Textures at 256x 48FPS with 1% at 21
Textures at 512x 50FPS with 1% at 23

I was shocked that the optimized 5950x faired no better. I was shocked that the 5800x3d was a lot better. Lastly, I find it interesting that really, the 1% lows don't change at all. All my 5800x3d experience for this was in my main system with a probably seven year install of windows 10 and all the pollution it has acquired where it was worse than the tuned 5950x.
The scenario you are testing: The game processes which aren't a "hit" for the cache, are probably first bandwidth/speed limited, by the DDR4. But you'd need a newer architecture anyway, to use DDR5. So you would double up on improvements.
 
The scenario you are testing: The game processes which aren't a "hit" for the cache, are probably first bandwidth/speed limited, by the DDR4. But you'd need a newer architecture anyway, to use DDR5. So you would double up on improvements.
Not im thinking a 7800x3d may be the best bang....

Frustrating...I thought I had it figured out.
 
Not im thinking a 7800x3d may be the best bang....

Frustrating...I thought I had it figured out.
Minecraft is wildly popular. There's gotta be footage of someone with a 4090 and in a similarly complex scenario, with newer CPUs: yeah?
 
Minecraft is wildly popular. There's gotta be footage of someone with a 4090 and in a similarly complex scenario, with newer CPUs: yeah?
It is the mods that make just about everyone's experience slightly different...

Heck there are lag generators you can build in a world to intentionally force fps to suffer.


I just want to get my ship done and sailing and looking nice in the distance of my world...been at this thing for months now.

That and i want to blow some stuff up.
 
It definitely is, they seem to be on sale currently.
now I"m looking at microcenter combos on amazon and then the builder on newegg. Thankfully, I do have ram, so I'm set there.

I just don't find any of the motherboards all that inspiring for the prices.

My wants aren't that tough...
I'd like to at least have it be pcie gen5 to be as future resistant as possible.
Not an asrock in case i want to change at some point in the future.
Has to have a x4 slot for a 10gbe nic.

the micocenter/amazon bundles have an msi 670e gaming plus wifi combo that is $600, but that seems like a lot when I can do a newegg one for $619 with an x870e tomahawk.

On newegg, the gigabyte boards don't have a spot for the nic or are more expensive than the tomahawk so seems a bit pointless.

That leaves the asus 650E max gaming wifi W, Prime x870-P wifi and x970 AYW Gaming wifi at $509, $569 and $579 respectively.
 
That leaves the asus 650E max gaming wifi W, Prime x870-P wifi and x970 AYW Gaming wifi at $509, $569 and $579 respectively.

I have the B650E Max Gaming Wifi W, I'm using it right now actually. Great board for the money, only thing that I dislike is it only has one flimsy m.2 heatsink but I use my own anyways. On this board the x4 slot is the very bottom one, I don't know about the others.
 
I have the B650E Max Gaming Wifi W, I'm using it right now actually. Great board for the money, only thing that I dislike is it only has one flimsy m.2 heatsink but I use my own anyways. On this board the x4 slot is the very bottom one, I don't know about the others.
I found some inspiration...

Amazon open box returns...

7800x3d and a x670e gaming plus wifi for $489.

They also have the x670e tomahawk and I could spend another $30 and I like the block layout better on the tomahawk, but I also like $30 and my ram is white so the gaming plus will be fine.
 
I found some inspiration...

Amazon open box returns...

7800x3d and a x670e gaming plus wifi for $489.

They also have the x670e tomahawk and I could spend another $30 and I like the block layout better on the tomahawk, but I also like $30 and my ram is white so the gaming plus will be fine.
If you’re only gaming the best price/performance/efficiency is the 7600x3d

It’s a microcenter exclusive at $230 but can be found second hand for like $225 used. I doubt you’re near an MC, but even if not you can possibly have someone do you a solid and purchase and mail one to you

7800x3d is indeed a great chip but a price bump up. Really up to you to determine if you want to keep spinning wheels or screwdrivers with a build.
 
Back
Top