Upgraded from 5800X to 7800X3D and my scores....don't make any sense?

M

MistaSparkul

2[H]4U
Joined
Jul 5, 2012
Messages
3,571
Not sure what's going on here? Game performance lines up though, 7800X3D completely SMASHES the 5800X hands down. I thought this would be a clear win for the 7800X3D but the CPU test is worst.

1680827666105.png
 
your gpu is 60mhz faster on the 5800x. are you controlling its speed with anything? maybe 3dmark needs an update? maybe not, it does look like it recognizes it properly.
 
pendragon1 said:
your gpu is 60mhz faster on the 5800x. are you controlling its speed with anything? maybe 3dmark needs an update? maybe not, it does look like it recognizes it properly.
Click to expand...

Yeah I wasn't expecting the Graphics score to be any different, and it makes sense given that it's running slightly lower clocks and thus ends up 1% slower. It's just the CPU part that I'm puzzled with. Only 3.9% better than 5800x in the CPU score and in the CPU test it's actually slower by 3.8%. Weird. I was expecting to reflect what I'm seeing in games in that the 7800X3D is a good 40% faster.
 
MistaSparkul said:
Yeah I wasn't expecting the Graphics score to be any different, and it makes sense given that it's running slightly lower clocks and thus ends up 1% slower. It's just the CPU part that I'm puzzled with. Only 3.9% better than 5800x in the CPU score and in the CPU test it's actually slower by 3.8%. Weird. I was expecting to reflect what I'm seeing in games in that the 7800X3D is a good 40% faster.
Click to expand...
then maybe an update to 3dmark is needed? actually, looks like there was a patch this morning. idk...
 
I mean it's not like I spend hours chasing scores or anything. I typically just run it one time whenever I get new hardware just to see what the improvement is, this time around the result was just a bit puzzling.
 
Try testing Cinebench and some other applications. Seems like something's unique with 3DMark
 
FlawleZ said:
Try testing Cinebench and some other applications. Seems like something's unique with 3DMark
Click to expand...

I believe Cinebench doesn't benefit at all from V Cache so the score would probably just be worst than the standard 7700X. In any case I guess it's just a 3DMark thing because this thing is totally smashing in games. Any CPU bottlenecks I had before on the 4090 even at 4K are totally gone like in Far Cry 6.
 
MistaSparkul said:
Not sure what's going on here? Game performance lines up though, 7800X3D completely SMASHES the 5800X hands down. I thought this would be a clear win for the 7800X3D but the CPU test is worst.

View attachment 562153
Click to expand...
What? It says the CPU score is 3.9% better with the 7800X3D compared to the 5800X. Only the graphics scores are lower, and that is due to the lower boost clock.
 
Armenius said:
What? It says the CPU score is 3.9% better with the 7800X3D compared to the 5800X. Only the graphics scores are lower, and that is due to the lower boost clock.
Click to expand...
hes saying that his in game perf is way better than what 3dmark is suggesting it would be.
 
Armenius said:
What? It says the CPU score is 3.9% better with the 7800X3D compared to the 5800X. Only the graphics scores are lower, and that is due to the lower boost clock.
Click to expand...

Yeah I wasn't concerned with the graphics scores. Just saying, 3.9% better in the CPU score but somehow 3.8% worst in the CPU test yet in games it's completely smashing the 5800X by far more than 3.9% when CPU limited. Typically synthetics tend to give the base case scenario and then real world gaming shows less of a performance gap but this time it's the opposite in that the real games showed crazy gains while the synthetic showed barely any improvement.
 
pendragon1 said:
then maybe an update to 3dmark is needed? actually, looks like there was a patch this morning. idk...
Click to expand...
New faster CPU comes out and you need patch to get higher CPU score than last CPU because benchmark didn't recognize CPU model to give it higher CPU score 🫣

Rev. Night said:
The gentleman above me do have excellent insights. That being said, the last time I cared about my 3dmark score was back in 2001. I remember a huge robot shooting missiles, and thats about it
Click to expand...
And this is exactly why most people stopped caring for 3dmark after 2001SE
It doesn't seem like benchmarks from this series properly test performance as much as they tests how much money companies are willing to put in to having higher 3dmark scores. They also usually have worse graphics than frame rate would indicate except maybe few exceptions.

Even 2001SE scores didn't make sense and seemed to be artificially limiting performance of non-TnL graphic cards just to make people think TnL gives such a big boost in performance. Games either ran almost the same with TnL or without it or later started to just refuse to run without it. Only 3dmark benchmarks showed this much performance difference. Also why when running without TnL - which by definition of it forces some calculations to be done by CPU - causes further increasing CPU speed more and more not make any performance difference at all?!?!
3dmark was always bogus benchmark, that's why.

Giant robot shooting missiles at the car was amazing though.
Also demo with 'nature' scene made huge impression on me the first time I saw it. From far away it looked almost photorealistic...
 
MistaSparkul said:
and in the CPU test it's actually slower by 3.8%
Click to expand...
No it isn't. It's 3.8% faster. Do not let the negative value fool you. Those CPU score numbers are frame times in milliseconds, not fps. Lower frame time=faster frame generation=higher fps. That's why it's green for the 7800X3D. It's a better score taking less time to generate a frame on the 7800X3D vs 5800X.
Armenius said:
What? It says the CPU score is 3.9% better with the 7800X3D compared to the 5800X. Only the graphics scores are lower, and that is due to the lower boost clock.
Click to expand...
Exactly.
 
Actually, you misread the CPU test. It's actually 3.8% better than the 5800X. The CPU test measures the amount of latency within the CPU, and lower numbers are better.
 
Ah ok got it. But yeah in actual games the differences are much bigger. Oh well I'm done worrying about 3DMark and have just been enjoying games on it now.
 
Haven't seen this thread talk about temperature yet, make sure your mount is good & not getting any thermal throttle. I get throttle with an ice giant on my 7950x. Also use latencymon and make sure no driver is wreaking your performance. I always install minimum required so I never bother with DPC latency checking anymore, unless I have audio popping. Good to look at tho if you want every ounce of performance before overclocking.
 
Thing about the 3DMark CPU test is that it's just that- a CPU test, not a gaming test.
The workload (based on utilization / power consumption etc) seems closer to something like Cinebench than any actual game, it's a maximally multithreaded workload that only measures compute throughout and basically ignores things like cache and latency.

For example, my Ryzen 5900X scores much higher than my 5800X3D did in 3DMark CPU simply because it has more cores. My 5800X scored higher than the 5800X3D because clock speed.

Even the most extremely multithreaded, compute-heavy game I know of (Cyberpunk with mods) doesn't behave like the 3DM CPU score would suggest.

Just enjoy the games and don't worry abt the 3DM CPU score, it means almost nothing for actual gaming with V-Cache CPUs in the mix.
 
I wish I could recall the exact phrase this site used when it used to do hardware reviews :)() because I LOVED the phrase. It went something like "we show these artificial benchmarks to show nothing is wrong, not to prove something is superior" (during CPU and GPU reviews).

Anyway, my point was, are seeing a degradation under real life work loads?
 
I know Nexus Gamer saw issues going from the 7950X3D to the 7800X3D due to the core parking drivers. They had to re-install OS to get that fixed. Not likely your issue but a fresh OS on a second spare drive might help and confirm if its a driver or old OS issue
 
You should be fine, its a margin of error at this point. I went ahead and ran it and compared to yours, but this also highlights how much power they shove into the 4090 because I'm running mine at 80% power limit to keep the fans quiet and I'm only 2.5% away from you. Undervolting ftw.

tse.png
 
Oh yeah I'm through thinking about 3DMark stuff and just been happily enjoying the new CPU. There are still plenty of games that are bringing this thing to it's knees even at 4K resolution, but I would say that's mostly been due to subpar optimization.
 
MistaSparkul said:
Oh yeah I'm through thinking about 3DMark stuff and just been happily enjoying the new CPU. There are still plenty of games that are bringing this thing to it's knees even at 4K resolution, but I would say that's mostly been due to subpar optimization.
Click to expand...

Sweet, I'm going to have one to play with in a day or two. Building a new rig for my friend and he wanted it, I have the board and RAM here already waiting. He hasn't chosen a GPU yet so I'm going to test it with my 4090.
 
You must log in or register to reply here.
Back
Top