nah just shut up now. everything you just said is addressed by using a save state. if the run to run variance was high they would have found a better way to test as they do multiple runs for each result and average
Cause there is no better way to test put it simply this is a ridiculous game to test for FPS. If you are telling me you are testing for CPU usage I’ll would still believe it. But for your information save state doesn’t mean anything in cities skylines 1 the game has a tendency if your traffic is to high they will despawn cars immediately. What if the demand change differently during opening the save? If one suddenly has a higher demand for high rise building and it change level? Will the traffic at the block of street increase? But if there is no demand for Low rise building will the building get abandon? Will the test gets abandon at the same time will it be abandon after the test? There is a huge variance to the test. I’m not suggesting the game is running fine it’s not it’s resource hogging the image for higher resolution and lower resolution is wired what I’m trying to say is if you want to test cities skylines don’t do it this way. This is like cooking food using the same ingredients from the same supermarket but putting ingredients using feeling and expecting it to have the same variance each time. I don’t know the variance is high or not and I don’t care. What I care is if they are doing the test properly, they cannot even keep thing in the game constant ,they can’t even get correct FPS for 1 run how do you think having multiple run will get correct FPS?
But I’m not doing YouTube video doing benchmarking like this guy I’m just run of the mil guy looking at Reddit thinking whether I have enough money to survive Tmr.I already suggested to force colossal order to come up with a testing mode like GTA V where everything will be the same the crowd density the same the traffic density the same and everything and let his test to his heart desire.
My job is not a hardware reviewer that tests game if thats my job I’ll will jolly well find a solution to this problem. It’s like being your member of parliament I’m not the constituency mp but if I feedback any issue I jolly well expect him to come up with a solution. If that’s your thinking I think everyone’s job would be easier .I quote “so you don’t have an actual suggestion, you’re just rambling” if that’s the case my member of parliament would be laughing to the bank with his salary as I need to suggest solution to the problem that I’m facing in my estate and not him coming up to solution.
nah just shut up now. everything you just said is addressed by using a save state. if the run to run variance was high they would have found a better way to test as they do multiple runs for each result and average
Cause there is no better way to test put it simply this is a ridiculous game to test for FPS. If you are telling me you are testing for CPU usage I’ll would still believe it. But for your information save state doesn’t mean anything in cities skylines 1 the game has a tendency if your traffic is to high they will despawn cars immediately. What if the demand change differently during opening the save? If one suddenly has a higher demand for high rise building and it change level? Will the traffic at the block of street increase? But if there is no demand for Low rise building will the building get abandon? Will the test gets abandon at the same time will it be abandon after the test? There is a huge variance to the test. I’m not suggesting the game is running fine it’s not it’s resource hogging the image for higher resolution and lower resolution is wired what I’m trying to say is if you want to test cities skylines don’t do it this way. This is like cooking food using the same ingredients from the same supermarket but putting ingredients using feeling and expecting it to have the same variance each time. I don’t know the variance is high or not and I don’t care. What I care is if they are doing the test properly, they cannot even keep thing in the game constant ,they can’t even get correct FPS for 1 run how do you think having multiple run will get correct FPS?
Stop with your pointless walls of text and suggest a better methodology then.
It needs to be repeatable, represent typical game performance and have lower run-to-run variance than GN’s methodology.
But I’m not doing YouTube video doing benchmarking like this guy I’m just run of the mil guy looking at Reddit thinking whether I have enough money to survive Tmr.I already suggested to force colossal order to come up with a testing mode like GTA V where everything will be the same the crowd density the same the traffic density the same and everything and let his test to his heart desire.
So you don’t have an actual suggestion, you’re just rambling.
I don’t think you understand GN’s methodology and how they ensure results are accurate.
My job is not a hardware reviewer that tests game if thats my job I’ll will jolly well find a solution to this problem. It’s like being your member of parliament I’m not the constituency mp but if I feedback any issue I jolly well expect him to come up with a solution. If that’s your thinking I think everyone’s job would be easier .I quote “so you don’t have an actual suggestion, you’re just rambling” if that’s the case my member of parliament would be laughing to the bank with his salary as I need to suggest solution to the problem that I’m facing in my estate and not him coming up to solution.