3DMark scores higher with lower clock speeds?

Lochli

Commendable
Feb 18, 2016
4
0
1,510
So, curiosity is pecking as I recently just upgraded my GPU for some extra horsepower on 1440p/X34 upgrade. Long time lurker (Years) first time poster...

For fun, I decided to tweak up my new card, and as I read the 3dmark HoF for the same test, Im seeing tons of lower clocked cards scoring +4000 points more then me, with lower clock speeds, and I'm curious: How is this possible? Is my older rig holding back the GPU?


Specs of rig:
i7 2600K @ 4.4
MSI p67a-gd65 MB
16gigs Gskill Ripjax X @ 1600
980ti KP 73%- Currently @ +115 on the GPU (1535ish?) OC. Havent started on the memory OC yet. Ran out of time. Temps never exceed 51C
1000w HX1000
1x 500gig Samsung Pro SSD
Many 120 Jetflow fans in an Antec 900
Windows 10 x64 Pro
Latest Gforce Drivers
 
overall score will be held back by CPU, but look at the graphics score. Thing is, I have found with 3DMark that there must be cheats...people getting 1000pts higher (gfx) than me with same GPU and clocks....something is odd!
 
Ya I figured the CPU might hold it back some..

But something doesn't seem right esp. when I look at ranks with the same GPU with a core clock of 1250 scoring 19000+ and mine with 1535+- clock and score of 15400, Im determined to see how this is possible.
 


3D Mark doesn't read clocks correctly most of the time. It's pretty normal. Especially if there's a custom bios being used.
 

3DMark doesn't always show the correct clock speeds, it doesn't for me when I overclock.
 
Interesting. People must cheat it though somehow...there aren't any clocks on earth that get my card to the scores posted...seems like the OP has the same issue.

shame as its a good comparison tool with lots of info.
 
Ya I figured the speeds wouldnt be percise, but, I didnt think they would be +-300mhz on cores. I mean kingpin #1 score shows a clock of 18/19xx whatever, Im sure they are pretty close to what hes actually using.

But when you drop 20-30 ranks, you see guys clocking in @ 1200+- with a +30% higher score, so either the lower ranks clocks are masked/not as accurate at the leaders, or something is going on here.
 

I use +217 which for me makes 1495Mhz, it depends on the stock clocks of that particular card.

 
I understand this, but with the boards showing the #1 guy, Kingpin utilizing a clock of 19/20xx whatever it is, leads me to believe it is actually reading the cards clock fairly accurate.

Which lead to my question, How is the SAME card with a IE.., 1200 core clock speed scoring almost 25% MORE than a card with a clock speed in the 1500+. Maybe the OC memory is making up the difference, but I highly doubt a 3/4000 point difference....



 


Are you talking about graphics scores, or overall scores? CPU plays a part in the overall scores. For just the graphics portion of the test, yes, VRAM clock makes a pretty big difference.

I've been using 3D Mark benchmarks for years, and it's a very rare occasion that they actually read the clocks right. With Keppler, they read the clocks right on the stock bios, but not on a custom bios. With Maxwell, they don't read right on the stock bios, but DO read right on the custom bios. I have a rig with an AMD GPU in it as well, and depending on which bios file I'm using, either stock or custom, will make it read funny as well.

Valley and Heaven report clocks wrong as well.

There's not some great conspiracy to keep you from making your way onto the Hall of Fame leaderboard. roflmao

 


Trick it? No...

Drivers? Yes....
 


What kinds of things are you talking about, exactly? You have a 390 and can't get it to clock high enough to score as high as some scores you've seen posted somewhere? Post your score here...what core clock are you running? What memory clock are you running? How much voltage? Have you flashed to a custom bios, or are you running the stock bios? Are you water cooling, or using the stock air cooling? What is the ambient room temperature?

There are a couple of hundred different things that can affect benchmark performance.....if I were to try to nail it down to one thing, I'd say it's very likely that the problem is PEBCAK....

 


They're notoriously slow getting the AMD drivers approved, btw....just something to keep an eye on. Especially if you're looking to get your score on the leaderboard for your hardware type.