Inconsistent score with exactly same setup (firestrike)

[Strategius]

I started using 3dMark recently and so far I could enjoyed it a bit. What I can't enjoy is that my score went down every time I did firestrike.
How can it be that my score went down after my first run of firestrike (every further run was lower than the one before)? I didn't change anything on my GPU and when comparing the values of the 3 runs you can see that it is only the GPU that decreases in value. I had a break inbetween to let the components cool down so that shouldn't be an issue. And my temps are just fine too (for a Laptop), my cpu never goes over 74°C. I didn't touch any settings related to the GPU. How can you explain this?

I mean that shows me that having the exact same clock speed and same settings isn't giving me same scores, which cannot be the fault of my hardware, can it??

1st run: https://www.3dmark.com/fs/17963252
2nd run: https://www.3dmark.com/fs/17963518
3rd run: https://www.3dmark.com/fs/17965375

I will do another run now and hope my score raises again.

As it is a laptop I can't really overclock due to overheating issues (only thing I did was untervoltage to cpu to lower temps on the cpu as I can't overclock the cpu anyways and even if, it would cause temp problems)

I am using a Acer Aspire 7 with a
GTX 1050ti with 4GB VRAM
i5 7300HQ
8GB DDR4 2400MHz
128GB SSD+ 1TB HDD
Can't find out which mainboard is in there

[Strategius]

4th run and score went down furhter... with the same clock speed of my GTX 1050ti, don't know what to do. https://www.3dmark.com/fs/17966828

[DCR]

It's a laptop, your results are perfectly normal.

[Strategius]

Thanks, I know I am happy with them, score and temp.
What I was actually referring to was not my score but the decrease after every run, which is not normal. You can expect at least almost the same score with the exact same settings and not a steadily decrease.

PS: After rebooting later and doing another run my score was higher again, so not sure if the first run is always the best and from then on it goes down, BUT WHY?

[DCR]

Time for you to accept it's a laptop. You shouldn't be running benchmarks on it. Enjoy it for the lifespan it has and treat it well and it may last you a long time.

[𝔇ave]

Originally posted by Strategius4:

Thanks, I know I am happy with them, score and temp.
What I was actually referring to was not my score but the decrease after every run, which is not normal. You can expect at least almost the same score with the exact same settings and not a steadily decrease.

PS: After rebooting later and doing another run my score was higher again, so not sure if the first run is always the best and from then on it goes down, BUT WHY?

It goes down because your laptop is getting hotter each consecutive run. This affects performance

[Strategius]

It goes down because your laptop is getting hotter each consecutive run. This affects performance

Thanks for the replay, I understand your point. On the other hand shouldn't the performance stay the same with exactly the same core/clock speed xxxxMHz, not matter if the temperatur goes up a few degres?

In other words 60°C - 3,1GHz and 78°C - 3,1GHz should give you the same performance (in my opininon)! I always monitor my core frequency and temps and I have no termal issues (never had) and I can see that my cpu runs at steady 3,1GHz with no jumps down to a lower frequency.

I would be happy and satisfied with your argument if the temp would cause my hardware to throttle down, but that is not the case at all. In fact my temps are almost the same every run with a marginal +-2-3°C difference. 3,1GHz are 3,1GHz why should they give you less performance with a bit higher temperatur, makes no sense to me?

I hope you get my point and I am just interested in this topic (benchmarking, speeds, temps etc.) in general and want someone who has some experience with benchmarking and can explains things to me. Thanks

[UL_Jarnis]

At higher temperatures, system starts to thermal throttle. So the clock speed falls. You can see this in Advanced Edition 3DMark hardware monitoring graphs. I'd bet that the clock speeds do not stay as high throughout the run on subsequent runs as thermal throttling happens.

[Strategius]

Thank you for the reply.

1st question:
I would like to know now: Can you assure that a identical CPU or GPU running at the exact same clock/core speed will score the same amount of points (given that it holds its frequency at a constant level)?

2nd question:
I am not only talking about my cpu now. I would like to know in general if for example Person A has a i5 6600k running totally stable at 4,4GHz and Person B has the same i5 6600k totally stable at 4,4GHz, should they get the same score? Otherwise it doesn't make any sense for me if you managed to OC your hardware to a certain point and are not being rewarded with the same score as a other person who managed to OC to the same point.

I hope you can help me answering these questions which will actually help me making my decision wheter to buy this software

[UL_Jarnis]

1. Yes, assuming Windows behind the scenes is not doing something different between the two systems. Since Windows is a multitasking OS, there is always inevitably some variance due to the OS - our official spec is that this should be no more than 3% between runs and in most cases it is less than 1%. This assumes no thermal throttling / automatic adjusting is happening by the hardware or the driver - if either happens, then all bets are off, whichever system is throttled more scores less.

2. Yes, assuming again Windows 10 / background programs are not eating up performance. There is a whole black-magic level sub-culture in "optimizing" windows for OC contest purposes... Stuff like disabling Windows Update by killing services. There is a long list of potential pitfalls when optimizing for the best possible score.... If Windows goes to check for updates during benchmark, that unsurprisingly can mean some performance loss. Or if Windows decides to start updating search index of the drive contents. Or if Cortana is running and one run gets input from the microphone to analyze while another doesn't. Or Antivirus application decides to update or do a scan. Or backup app starting a backup during benchmark. Or....

If you carefully eliminate all possible sources of variance in available system resources to the benchmark, then the result will be very repeatable.