• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Official Neogaf 3DMark Timespy league table - the ULTIMATE epeen showdown

PaintTinJr

Member
The ram I bought was separate
MBQQQk4.jpeg


I think I chose correctly, CL30 6000 mt
If it is these ones


I'd recommend double checking they do support EXPO (AMD's XMP) boost profile, as I'm not seeing it on that page - unless I'm totally missing it - but I probably looked because I'm looking at a system upgrade(AMD or Intel looking at some scores) and wasn't sure I'd seen better than CL32 for EXPO.

edit: I seem to recall one of my friends in the last 12months tell me he'd sent back memory either because it didn't support EXPO, or because the timings were for XMP, and the EXPO timings were different and less impressive.
 
Last edited:
I did the benchmark with a bunch of apps open and half my ram was being used by firefox. I reckon I could squeeze a few more points out of it but honestly I don't care, and I'm happy if AMD actually begins competing with Nvidia.
Your CPU temps are a bit fiery though, 85º is the absolute maximum I've seen mine achieve.
Bought a new cooler, made a small difference to the score so there was some CPU throttling going on. Overall temps much better and the fan itself is almost silent. I went with the Arctic Freezer 36.

TlfH6h8.png


I'll try undervolting the 9070XT sometime. Maybe I can push over 25k.
 
Last edited:

skneogaf

Member
If it is these ones


I'd recommend double checking they do support EXPO (AMD's XMP) boost profile, as I'm not seeing it on that page - unless I'm totally missing it - but I probably looked because I'm looking at a system upgrade(AMD or Intel looking at some scores) and wasn't sure I'd seen better than CL32 for EXPO.

edit: I seem to recall one of my friends in the last 12months tell me he'd sent back memory either because it didn't support EXPO, or because the timings were for XMP, and the EXPO timings were different and less impressive.

This was the ones I've ordered, it does say optimised for AMD EXPO



Found it copy and pasting the Manufacturer code: CMH64GX5M2B6000Z30


 
Last edited:
Time Spy - The "E" cores arent useful in normal games, but it seems that the Time Spy results depends on them.The i9 14900 is almost twice as fast compared to my 7800X3D in this CPU test.



3333.jpg


Fire Strike Ultra


Ray Tracing Test (PT) 93.44 FPS


Port Royal


Speed Way


Steel Nomad

Yes the 3DMark CPU test is bunk in 2025 when comparing Intel vs AMD for gaming
 
Last edited:

PaintTinJr

Member
Yes the 3DMark CPU test is bunk in 2025 when comparing Intel vs AMD for gaming
It is possible game benchmarks on Windows with AMD are merely doing better because of the console having had AMD software for two generations so AMD optimisation has become the first priority given the gains are multiplatform, and Intel optimisation is maybe less by developers.

An easy check would be to take the biggest delta performing gaming between 9800X3D and Core Ultra 9 and benchmark on Linux using proton so that optimisation would have parity, and then see if the delta gets bigger or smaller. If the gap narrows it might suggest this benchmark is closer indicator of relative hardware CPU performance than current games.
 
Just a tad I may have been able to get more out of it but after running the same benchmark for over an hour and a half I got bored. :pie_roffles:
If that's a tad then maybe a 32k graphics score is possible on my Sapphire Nitro. Will definitely give undervolting a go soon, maybe I'll run Steel Nomad on loop until the driver crashes.
 

L0rdMike

Neo Member
If that's a tad then maybe a 32k graphics score is possible on my Sapphire Nitro. Will definitely give undervolting a go soon, maybe I'll run Steel Nomad on loop until the driver crashes.
Sorry that was sarcasm, I think it was a pretty extreme undervolt and was amazed it passed the tests. I dont know what the standard voltage is on a 9070XT but the 7900XT is 1100mV and I had mine at 985mV
 
It is possible game benchmarks on Windows with AMD are merely doing better because of the console having had AMD software for two generations so AMD optimisation has become the first priority given the gains are multiplatform, and Intel optimisation is maybe less by developers.

An easy check would be to take the biggest delta performing gaming between 9800X3D and Core Ultra 9 and benchmark on Linux using proton so that optimisation would have parity, and then see if the delta gets bigger or smaller. If the gap narrows it might suggest this benchmark is closer indicator of relative hardware CPU performance than current games.
AMD is doing worse, and it's because of the invention of 3D CPU's

Intel CPU's are only 2D but they have a larger core count due to the P + E cores layout, and the 3DMark CPU test greatly benefits from more cores even if the cores are E cores which are useless in gaming

The AMD CPU's have fewer cores because they never implemented P + E, all AMD cores are effectively P cores, but they are 3D and the massive Vcache greatly improves gaming performance

This is why the 5800X3D is such a meme CPU, because while it has a comparatively low clock speed and is on an obsolete architecture, it has surprisingly good performance in many games because of the Vcache
 

M1987

Member
Time Spy - The "E" cores arent useful in normal games, but it seems that the Time Spy results depends on them.The i9 14900 is almost twice as fast compared to my 7800X3D in this CPU test.



3333.jpg


Fire Strike Ultra


Ray Tracing Test (PT) 93.44 FPS


Port Royal


Speed Way


Steel Nomad

It's obviously nowhere near in RT,but It's a bit weird I got a better Port Royal score with a 9070XT and 7600x
 
It's obviously nowhere near in RT,but It's a bit weird I got a better Port Royal score with a 9070XT and 7600x
The RX 9070XT can run Port Royal (hybrid RT test) very well, 1fps better than RTX4080S in guru3d benchmark.


In real games 9070 XT results in hybrid RT games are all over the place though. In some games the 9070XT is close (within 20%) compared to my card, but there are also games, where my card is almost twice as fast. In PT that difference is even bigger, up to 211%.
 
Last edited:

PaintTinJr

Member
AMD is doing worse, and it's because of the invention of 3D CPU's

Intel CPU's are only 2D but they have a larger core count due to the P + E cores layout, and the 3DMark CPU test greatly benefits from more cores even if the cores are E cores which are useless in gaming

The AMD CPU's have fewer cores because they never implemented P + E, all AMD cores are effectively P cores, but they are 3D and the massive Vcache greatly improves gaming performance

This is why the 5800X3D is such a meme CPU, because while it has a comparatively low clock speed and is on an obsolete architecture, it has surprisingly good performance in many games because of the Vcache
As far as I understand it, 3dmark is still tied to the Intel thread director assigning work to cores and the benchmark doesn't provide fake busy work for e-cores, just has the render/simulation tasks highly threaded so they can use whatever is available.

So having had a little chat with co-pilot it maintains the scores are largely based because of the the threading and 700Mhz per P-core (x8) advantage the Intel chips have over the AMD ones, and that the e-cores aren't being used for lower cadence gaming tasks, like geometry sorting per second, visible determination of geometry containers using octrees or L3 and RAM garbage collection, etc, so I'm not entirely convinced the benchmark is getting it totally wrong for Intel vs AMD.
 
Last edited:

M1987

Member
The RX 9070XT can run Port Royal (hybrid RT test) very well, 1fps better than RTX4080S in guru3d benchmark.


In real games 9070 XT results in hybrid RT games are all over the place though. In some games the 9070XT is close (within 20%) compared to my card, but there are also games, where my card is almost twice as fast. In PT that difference is even bigger, up to 211%.
Hopefully UDNA can catch up in PT like they have in upscaling and the improvement in RT on the 9070 series
 
Last edited:

Fyrium

Neo Member
Did this for giggles since I'm sure it won't rank very high. On the plus side, I now have 3dmark if I ever need to test things.

Fyrium Fyrium - Time Spy - 20657

I'm sure it can do better somewhere, but I'll probably just be upgrading to AM5 later in the year and moving to 32GB ram.
 
Top Bottom