Question Is the 3DMark comparison score wrong ?

[ridwan123]

I always compare 3DMark score between each GPU when i decided to buy new GPU and why when i see benchmark on youtube it is not accurate ?

Like, for example, 3DMark score for 4070 ti super is 24000 in time spy and 7900xt is 26000 in time spy normal and why during benchmark the 4070 ti super is somehow faster than 7900 xt?

 

[CountMike]

I always compare 3DMark score between each GPU when i decided to buy new GPU and why when i see benchmark on youtube it is not accurate ?

Like, for example, 3DMark score for 4070 ti super is 24000 in time spy and 7900xt is 26000 in time spy normal and why during benchmark the 4070 ti super is somehow faster than 7900 xt?

Benchmarks are difficult to compare unless under exact circumstances and identical setups. Everything even temperatures can affect scores let alone CPU, RAM etc. It's not practical to base your decision on one benchmark, best to look at real life performance on game to game basis.

 

[ridwan123]

Benchmarks are difficult to compare unless under exact circumstances and identical setups. Everything even temperatures can affect scores let alone CPU, RAM etc. It's not practical to base your decision on one benchmark, best to look at real life performance on game to game basis.

As i know RAM,SSD,CPU dont matter in gpu benchmark in 3dmark so there will be no bottleneck at all.it only measure pure gpu perfomance only .it is different from games where cpu matter

 

[CountMike]

As i know RAM,SSD,CPU dont matter in gpu benchmark in 3dmark so there will be no bottleneck at all.it only measure pure gpu perfomance only .it is different from games where cpu matter

Can't have GPU without CPU and RAM so they do figure in GPU performance.

 

[kerberos_20]

As i know RAM,SSD,CPU dont matter in gpu benchmark in 3dmark so there will be no bottleneck at all.it only measure pure gpu perfomance only .it is different from games where cpu matter

ram and cpu does affect gpu benchmark
or do you think gpu can work without CPU?
what about gpu drivers, do you think it runs without cpu?
then why is there like single core load when running gpu benchmark,
does DX api translation runs without CPU?
do you think that reducing ram latency would make your cpu run faster in that single threaded workload, since it would have lower wait times before moving data from ram to vram?

 

[ridwan123]

ram and cpu does affect gpu benchmark
or do you think gpu can work without CPU?
what about gpu drivers, do you think it runs without cpu?
then why is there like single core load when running gpu benchmark,
does DX api translation runs without CPU?
do you think that reducing ram latency would make your cpu run faster in that single threaded workload, since it would have lower wait times before moving data from ram to vram?

Of course gpu cannot work without cpu.what i mean in 3dmark your gpu score would not decrease with lower tier cpu but still it cannot run without cpu.

Mod Edit:

As you know i got i7 14700f processor paired with 4070 super zotac twin edge oc white edition and i got z790 motherboard paired with 6000 mhz Ram 16×2 and 850 w psu 80+gold

 

[CountMike]

Of course gpu cannot work without cpu.what i mean in 3dmark your gpu score would not decrease with lower tier cpu but still it cannot run without cpu

Those benchmark programs simulate games and results are for whole graphic system not just GPU itself.

 

[kerberos_20]

well anyway, back to your original question, 3dmark score on website is probably some random number, its not even based on average score from all theirs benchmarks..meaning from lowest low power/overheated scores maybe running vsync to highest overclocked benchs

now lets do some small comparison with mine GPU for example
RX 6800 based on this got 16258 gpu score points
https://benchmarks.ul.com/hardware/gpu/AMD+Radeon+RX+6800/review

now if i run timespy, i get this result: (18.4k gpu points)

I scored 17 152 in Time Spy

AMD Ryzen 7 5800X3D, AMD Radeon RX 6800 x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

from within app theres also graph showcasing how does it compare with people with same hardware

https://imgur.com/4iuFBly

View: https://imgur.com/4iuFBly


gpu in question is msi gaming X trio, meaning overclocked edition with beefy VRM and cooler

so, you get what you pay for, if you buy low end model (bad bin needing more voltage per clock, bad cooling/vrm), youll endup with low score (if benchmark score what youre really looking for)

 

[thestryker]

Of course gpu cannot work without cpu.what i mean in 3dmark your gpu score would not decrease with lower tier cpu but still it cannot run without cpu

This isn't quite accurate the score does decrease with a slower CPU, but not dramatically so.

Like, for example, 3DMark score for 4070 ti super is 24000 in time spy and 7900xt is 26000 in time spy normal and why during benchmark the 4070 ti super is somehow faster than 7900 xt?

If you look at 3dmark scores they're absolutely a calculated average so you would need to directly compare rather than take whatever the expected value is. The best thing 3dmark is good for is the results database and directly comparing similar hardware so you can ensure your performance is roughly where it should be.

GPUs also perform differently depending on what the task is which is why reviewers use a bunch of different games that have different performance profiles to test video cards with.

 

[ridwan123]

Can't have GPU without CPU and RAM so they do figure in GPU performance.

well anyway, back to your original question, 3dmark score on website is probably some random number, its not even based on average score from all theirs benchmarks..meaning from lowest low power/overheated scores maybe running vsync to highest overclocked benchs

now lets do some small comparison with mine GPU for example
RX 6800 based on this got 16258 gpu score points
https://benchmarks.ul.com/hardware/gpu/AMD+Radeon+RX+6800/review

now if i run timespy, i get this result: (18.4k gpu points)

I scored 17 152 in Time Spy

AMD Ryzen 7 5800X3D, AMD Radeon RX 6800 x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

from within app theres also graph showcasing how does it compare with people with same hardware

https://imgur.com/4iuFBly

View: https://imgur.com/4iuFBly


gpu in question is msi gaming X trio, meaning overclocked edition with beefy VRM and cooler

so, you get what you pay for, if you buy low end model (bad bin needing more voltage per clock, bad cooling/vrm), youll endup with low score (if benchmark score what youre really looking for)

I Only got zotac brand which is the low price at 4070 super maybe it just need optimization like asus and gigabyte which has higher price.maybe need to update v bios or having a good brand in the future.i only care about the perfomance in game but doubt when the result in 3dmark and passmark is different when it comes to AAA game

 

[ridwan123]

Those benchmark programs simulate games and results are for whole graphic system not just GPU itself.

Hmm

 

[ridwan123]

This isn't quite accurate the score does decrease with a slower CPU, but not dramatically so.

If you look at 3dmark scores they're absolutely a calculated average so you would need to directly compare rather than take whatever the expected value is. The best thing 3dmark is good for is the results database and directly comparing similar hardware so you can ensure your performance is roughly where it should be.

GPUs also perform differently depending on what the task is which is why reviewers use a bunch of different games that have different performance profiles to test video cards with.

Because when i need to buy gpu in the future i always see 3dmark result and passmark and if the result is different in actual games it become complicated

 

[kerberos_20]

Because when i need to buy gpu in the future i always see 3dmark result and passmark and if the result is different in actual games it become complicated

than you would need to look for specific GPU brand/model, see if it has review done
you can also search for PCB and compare with different models (VRM vs available power)
you can also check vbios for that GPU on techpowerup vbios collection, that should show gpu/memory boost clocks and power limits (higher power is better for performance as gpu can boost higher when needed)

 

[CountMike]

Best use for benchmarks is when you are tuning system, making changes/upgrades, updating drivers or suspect something wrong. Save results and compare with new ones. Like that you are staying with same parameters and know if you are advancing o regressing.

 

[ridwan123]

Best use for benchmarks is when you are tuning system, making changes/upgrades, updating drivers or suspect something wrong. Save results and compare with new ones. Like that you are staying with same parameters and know if you are advancing o regressing.

Maybe i do it