• 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.

Fallout 4 CPU benchmark thread(need some Zen3 and Zen4 results!)

New post over on AT:
User yuke improved their rankings from #25 to #2. Mostly be swapping from a 7900 XTX to an RTX 2080:
vPN5epA.png

DX12 is one thing, but for DX11 it seems that for FO4 the Radeon drivers have an overhead problem.
A bit of a concern for me as I still play the game. I hope the Anniversary Edition or whatever it is called is better optimised!
 
Opps, sorry. Copy and paste error, almost the same indeed!
Corrected it above, and here's the top 10:
71y9ten.png

So that's better. Number 4, probably without BLK overclocking like yuke and even faster memory it is hard to get much higher.

I keep including the S1/S2 ratio in the hope that it tells us something about drawcall vs NPCs (as CoreVega should have far more NPCs), but while there is variation I am unsure it means much.

It's outside and most of the NPCs are spread out in a large area - Diamond City has a lot of NPCs in a relative small area.
 
Yes, that's a pretty poor showing for such a power hungry system. At least you should have plenty of PCIe slots for multiple NVMe etc. (all those 2TB being cheap now but being only able to have a few NVMe slots - even with using a PCIe slot for another - means I am still holding on for 4TB to get cheap!).
iAuHheh.png

So about the same as that Zen APU with onboard (not that these benches test the GPU much but the iGPU will rob the CPU of memory bandwidth) with slightly more power usage.
Sweclockers tested the Intel HEDT chips of that generation and Fallout 4 didn't like them. Must be the Mesh topology they use.
 
@KompuKare

Made a new thread WRT to CPU performance in Starfield:
 
Last edited:
Without someone with the hardware and time and skill to totally trace and service this, we will probably never know. AMD engineers could probably have done so but sponsorship doesn't seem to mean they do anything deep!

Could be anything:
from a compiler flag - but Bethesda no longer leave things in x87 fall back mode!
Maybe even a single instruction Intel does better with and which is used a lot by, e.g., the Papyrus runtime.
Or some branch where Intel's branch predictor is better.

Just hope that whatever it is doesn't get fixed to prevent some speculative execution exploit!

I edited my post - I put CPU benchmarks in a new thread.
 
GameGPU tested the new update:

The Nvidia dGPU bias seems to still be there,but AMD CPUs seem closer to the Intel one?

@KompuKare Time to start a new benchmarking thread with the update?
 
That's what is so cool about these forum threads. Literally nowhere else have I seen draw call performance measured in an actual game that people play that can get complex enough to really push the limit. Like all the other Fallout 4 reviews just do a 5 minute fly through of Sanctuary and Concord. Which obv makes the numbers look really nice compared to what you get after actually making a settlement.

Oblivion can make modern CPUs cry with how ancient the code is. I remember enabling LOD models for every piece of architecture, and being royally disappointed by my new i7 6700k lmao. Was barely pushing to 13fps depending on the camera angle because of all the draw calls.

I'm really curious to see how intel's draw calls are. Because if they are actually better than AMD they might actually be worth buying for that.

Only one review site bothered doing settlement benchmarks and that was HardOCP. Some of the Polish websites tested parts of the cities which also seem to have push a heavier load. The other benchmarks were too light. But even with your Core i7 fast DDR4 would make a noticeable difference over slower RAM.

The reason I pushed this thread over on this forum,was a lack of modern CPU test results in the AT thread. Hopefully when the Zen5 CPUs are out we can check the results.
 
Last edited:
While we wait for Zen 5 Ryzen (especially 9800X3D) and ADL results, I though I'd see how Fallout 2024 (v1.10.984.0) performs like vs previous (v1.10.163.0).
I recently updated to a used RX 6800 from my RTX 3050 which should see a bit of a regression here (but hey, 16GB vs 8GB is great for Skyrim modlists at 1440P with ones like Nolvus stating 12GB+ for 1440P).

A40UGtB.png


Now, 55.9 / 57.9 is about 96.5% so a loss of 3.5% or so.

Would need a larger sample size and more CPUs to be give a more definite conclusion.

I still haven't benchmarked my 7800X3D because I am playing Fallout:London ATM.
 
@KompuKare Had a look through the thread and found you had uploaded them:

Okay, pCloud it is then:
SHA256 should be:
f2d09283642b85a4af11580adcefaf850e26e0f9fdc1f94ace5b8316b51225ba
Those are the files I send to @themontanio in post #342 above. Ini and saves are the ones we should use. For the ENB version there may be a newer one, but I don't think it matters much.
 
Did a few quick tests using my old install. Might get a bit better FPS with a fresh install.

Common hardware:
1.)Ryzen 7 7800X3D set to 65W Eco mode with undervolting
2.)32GB 6000CL32 RAM

GPU:RTX3060TI
Driver:566.36
Corvega:125.3/122.8/123.6FPS(Average:123.9FPS)
Diamond City:140.5/136.5/142.4FPS(Average:139.8FPS)

GPU:RX9070
Driver:25.3.1
Corvega:92.5/92.8/94.8FPS(Average:93.37FPS)
Diamond City:113/110.5/110.9FPS(Average:111.47FPS)
 
The RX9070XT was around £50 more and I got what I could get. In terms of price/performance it's not massive ahead - around 10% plus the need for a bigger PSU. The performance/watt of the RX9070XT is not great and the coolers are massive. Even the Pulse is enormous despite being one of the smaller cards - it's about 1.5X the size of my RTX3060TI FE despite only 20W more board power. It's also larger than my RX5600XT Pulse.
CrghHod.jpeg

JI9VSWc.jpeg


irWLWez.jpeg



TBF,the second rung cards interested me more:
1.)9500 PRO instead of 9700 PRO
2.)7900GS instead of 7900GT
3.)X1900
3.)HD4830 of HD4850/HD4870
4.)HD5850 instead of HD5870
5.)RX470 instead of RX480
6.)RTX3060TI instead of RTX3070

I really don't like cards over 300W,or even 250W due to heat and noise.

The only RX9070XT I was after was the Powercolor Reaper RX9070XT and it had half the RRP stock of the Sapphire RX9070XT Pulse. My current case cannot take a 3 slot card unless very small,so the only cards at RRP which I could get which would fit were:
1.)The Sapphire RX9070 Pulse which is the shortest RX9070 series card but 2.5 slots
2.)The Powercolor RX9070/RX9070XT Pulse which are longer but 2 slots

However,I could only get an RX9070 Pulse at RRP,not any of the others. I couldn't get a Powercolor card at 2pm but managed to put the Pulse in my basket and it eventually took me 2.5 hours to complete the purchase.

OcUK dropped a few cards at 10pm,but I was not sure I could buy two RRP cards and the stock lasted two to three minutes. I almost managed to the RX9070XT Reaper but decided not to bother.

Moreover,I wouldn't want to run an RX9070XT off an SF600 PSU. I already know,how much my RTX3060TI system runs at the wall(around 300W to 330W) with an Ryzen 7 5700X. Someone here with a Powercolor RX9070XT Hellhound and a Ryzen 7 5700X said they got a power draw of 426W at the wall!!

The PSU would be running at 2/3 it's rated capacity and the RX9070 cards do have spikier power looking at the reviews.

The reality is I would need to get an SF750/SF850 and that would be around £150 to £170.

Also undervolting the RX9070XT would need me to drop power closer to 250W and the issue,it's hard to say whether it would be massively ahead of the RX9070 at that sort of power. The issue is if got a card which didn't undervolt well,I would need to power limit the card which would drop performance.

The RX9070 could be pushed upto that power and also tweaked.

Then there is all the heat and noise too.


Thats...strange results?

Nvidia cards run this game better.
 
Last edited:
@KompuKare What does the entire list like so I could update the OP?

My results might be not the best because my install is a Fallout:London install with some mods installed,which I jerryrigged for this test.

Also my Ryzen 7 7800X3D is in eco-mode with an undervolted and I could get 6400MHZ RAM to run stably for more than a few weeks,so just whacked it down to 6000MHZ with no tweaking.

When I have some time,I could try a clean install of Fallout 4 to see what performance is like.

In some other games like Monster Hunter World,I have over a doubling of performance.
 
Last edited:
New entrant over on the AT thread on this:
They got the #1 in the rankings:
OryMwvk.png

Their Diamond City score is only the 2nd best, but their Corvega score is far ahead of every other score. I suspect Diamond City with all the NPC is less drawcall focused and hence the Corvega results imply they have a huge lead in drawcalls there!

Complete list - this is now so long that at a reasonable zoom level it takes over two 1400P screens.
fEUba1m.png

Still no 9950X3D results the 7800X3D vs 7950X3D result don't really show more cores helping.
I will try and get a RX9070 result at some point with a 64GB kit I acquired.
 
Last edited:
Back
Top Bottom