-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Testing CPU bottlenecks using multiple CPUs and a GTX 1080ti
- Thread starter (G{in}[AK)TION]
- Start date
OP
![(G{in}[AK)TION]](/forums/data/avatars/m/116/116613.jpg?1469679448){kind=avatar}
- Joined
- Apr 14, 2011
- Thread Starter
- #102
That is true Kenrou. But like some people nowadays if not most, i am using an SSD while running my games from an HDD. I also have a bright HDD led light installed on the board that is on the test bed so i know that if i experience a stutter and i see it happen right wen the light comes on, then ill know when the hdd is being a bottle neck. But from what i saw in my testing, it rarely blinked during gameplay.
I have 16GB of ram installed on the 3770k and 8gb of ram installed on the 7700k. the 7700k was also running windows on an SSD while the games were loaded from the hard drive. So i doubt the page file could be a problem here if it ever was.
I have 16GB of ram installed on the 3770k and 8gb of ram installed on the 7700k. the 7700k was also running windows on an SSD while the games were loaded from the hard drive. So i doubt the page file could be a problem here if it ever was.
FCAT measures frame times. But again, not sure that is a rabbit hole that should be travelled with the testing in its current state. Too many variables causing the data to be too general (at best).
Suttering can happen with a HDD, SATA based SSD, or even a PCIe NVMe based SSD gents. Both are exponentially slower than RAM. It simply depends on how the game works. Obviously MMORPG's, in general, beat on the HDD more than most FPS games so there is that to consider as well.
Personally, I think this needs to be looked at again with only using SP games and preferably with integrated benchmarks. MP testing isn't done for a reason (I've said that likely half a dozen times already
). We started off looking for one thing, bounced to another, and now a third idea when the fundamentals of the first set of testing isn't in order. I like the zeal you are showing, but, it the testing should be done more empircally. This is why every ****, tom, and jane, doesn't review because it takes more than to just run some benchmarks are compile the data. It isn't rocket science (clearly if I am doing it, LOL!) but it takes more than just being able to run things and collect data.
Last edited:
- Joined
- Dec 19, 2012
My point was, that on certain benchmarks take Rise of the TomRaider for instance when testing maxed out at 1440 I get stutters in the same spot all the time because the minimum frame rate has dropped way below 60 FPS. As Kenrou said there can be many causes but most often in my case, it's just the minimum frame rate hit's it's low at say 21 FPS and causes the stutter. This is with a 1080Ti even. This happens whether I'm using AMD or Intel.
OP
- Joined
- Apr 14, 2011
- Thread Starter
- #105
FCAT measures frame times. But again, not sure that is a rabbit hole that should be travelled with the testing in its current state. 1. Too many variables causing the data to be too general (at best).
2. Suttering can happen with a HDD, SATA based SSD, or even a PCIe NVMe based SSD gents. Both are exponentially slower than RAM. It simply depends on how the game works. Obviously MMORPG's, in general, beat on the HDD more than most FPS games so there is that to consider as well.
3. Personally, I think this needs to be looked at again with only using SP games and preferably with integrated benchmarks. MP testing isn't done for a reason (I've said that likely half a dozen times already
4. I like the zeal you are showing, but, it the testing should be done more empircally. This is why every ****, tom, and jane, doesn't review because it takes more than to just run some benchmarks are compile the data. It isn't rocket science (clearly if I am doing it, LOL!) but it takes more than just being able to run things and collect data.
1. What do you mean too general?
2. Also true, but with an SSD (from what i have seen at least) it would happen less often. off the top of my head, I remember seeing alot of stutter in Rez:Infinite and Battlefield 4 when run straight off the HDD. With an SSD, it was rarely ever there. But i play FPS games not MMORPGs. And the 7700k and Q9650 were the only ones that had 8gb of ram while the 3770k, fx-8350, and 7900x had 16gb of ram. I did get another DDR4 3200 16gb ram kit in the mail just a few days ago (Quad Channel reasons) so i will be testing again on the 7700k and 7900x just to make extra sure im not missing something here.
3. (and i have responded half dozen times already
) that i tested both single player and multiplayer. I mean, its gotten me under the impression that you dont like seeing them on the chart. Ouo *idk Shrugs*. Yes that is a smiley face with one eye being larger than the other. But you are right. I do think that i was bouncing around and when i did realize it, i tried to "keep the train on the right tracks but still taking the wrong turns".4. Enlighten me then, exactly how would you run these tests if you were in my shoes here? I am genuinely curious. I mean all i did was make sure the game was at "ultra" or "very high" graphic settings, run the tests in each game more than twice by playing the same chapter or level with fraps recording FPS (using the benchmark binded to f11 key), compared results making sure they werent too far apart, and then added them to the spread sheet. I am feeling like everything i did up until now was done incorrectly and it was all a waste of time.
Also, true. But i am trying to refer to the kinds of stutters that can happen any time during gameplay. Like windows decides to do something in background or because your doing this certain particular action in game or just the kinds where it happens for some random unexplained reason even if there was 16gb of ram or more in the system. But now i also feel like this should be done as a seperate test with the question "Does more cores help with keeping a more consistently smooth gameplay"
Last edited:
1. As I have alluded to (or said) all along in the thread, the MP tests are junk. There are too many variables which can affect the FPS as well as the gaming experience. SP run throughs, while not junk, are certainly not optimal.
2. It would happen just as often, but perhaps with less length of stuttering. If a texture/scene needs recalled from the HDD, it needs recalled from an SSD as well. The difference being the SSD is faster and, in theory, would show less stuttering.
2a. You could be hitting 8GB RAM use in those other systems. I recall asking earlier if you confirmed you were not hitting close to 8GB, but do not recall a response. So long as the titles you are testing aren't close to 8GB use, then you are OK. You are not missing anything going to quad channel RAM (reviews are on the web).
3. I don't like MP testiing period for the reasons I listed already. But, you insist on testing it like its giving good data.
3a. I also recall asking if you ran the MP tests again to see how much different the FPS were and do not recall a response. They may not be off much, but a couple % here and there adds up.
4. I've mentioned already how I would run the tests.
No MP games. Use SP games that have integrated benchmarks
Again, being a reviewer for so long, I am used to testing being empirical and repeatable which a lot of this testing doesn't appear to cover. The information being relayed, while it has some value, isn't teribly repeatable (the MP tests) and your SP tests are ALL based on manual run throughs. If I can be frank, about the only data set worth it to me is the one benchmark that isn't MP and has the integrated benchmark (Bioshock: Infinite).
2. It would happen just as often, but perhaps with less length of stuttering. If a texture/scene needs recalled from the HDD, it needs recalled from an SSD as well. The difference being the SSD is faster and, in theory, would show less stuttering.
2a. You could be hitting 8GB RAM use in those other systems. I recall asking earlier if you confirmed you were not hitting close to 8GB, but do not recall a response. So long as the titles you are testing aren't close to 8GB use, then you are OK. You are not missing anything going to quad channel RAM (reviews are on the web).
3. I don't like MP testiing period for the reasons I listed already. But, you insist on testing it like its giving good data.
3a. I also recall asking if you ran the MP tests again to see how much different the FPS were and do not recall a response. They may not be off much, but a couple % here and there adds up.
4. I've mentioned already how I would run the tests.
No MP games. Use SP games that have integrated benchmarks
How are you going to test for something that isn't repeatable? More 'butt dyno' results?? "Well, I don't *think* it stuttered as much...
Again, being a reviewer for so long, I am used to testing being empirical and repeatable which a lot of this testing doesn't appear to cover. The information being relayed, while it has some value, isn't teribly repeatable (the MP tests) and your SP tests are ALL based on manual run throughs. If I can be frank, about the only data set worth it to me is the one benchmark that isn't MP and has the integrated benchmark (Bioshock: Infinite).
Last edited:
@(G{in}[AK)TION]
Did you run the multiplayer games over again to see if you achieved consistency? Don't remove the multiplayer it is my proof that they use up more CPU cycles than single player AI.
Did you run the multiplayer games over again to see if you achieved consistency? Don't remove the multiplayer it is my proof that they use up more CPU cycles than single player AI.
Last edited:
Just two quick run throughs from PUBG through FRAPS... same level, landed at the same place on the map, had two engagements. A 5%+ difference in these two back to back runs. Not sure how other games will translate, but I assume in a similar manner.
OP
- Joined
- Apr 14, 2011
- Thread Starter
- #109
1. It is fine that you think they are junk. I been saying this whole time that i am testing both SP and MP because i am curious about both sides. SP run through results are pretty much, to me, would be a valid way of testing. I mean where can i find battlefields SP integrated benchmark? or Dooms? Rocket Leagues? Crysis 3?
2. I thought i did post my response but i dont see it here in the thread. I try to type my messages carefully (because of past life lessons) and if it takes too long for me to type then i close the window and come back later. But here is my response, Yes it was hitting the full 8GB and i will test again later on today. Video stuttering topic is now a lost cause to me because if i cannot fully explain what i am experiencing/seeing with it, then forget i ever asked.
3.I will take this as opinion. And i also failed to respond to your BF multiplayer question: Yes the results were similar. About 2-5Fps difference. Same map on same 60hz server. This also goes for the other MP games i tested. similar results.
4. Bioshock infinite and sleeping dogs are the only games i know of that have integrated benchmarks in my games list. the only reason why i did not post sleeping dogs Integrated benchmark was because i found out about it after i got to the 7700k when i was checking settings it was right on the bottom of the screen and i missed it. But i went ahead posting my SP run through thinking that it would show a valid result which to me, it did after 2 of the runs i did. I know that Counter Strike Global offensive also has a benchmark tool available on steam workshop but i could not figure out how to run it. I am sure that i posted somewhere stating this but either i have a poor attention span or no one actually responded to it.
2. I thought i did post my response but i dont see it here in the thread. I try to type my messages carefully (because of past life lessons) and if it takes too long for me to type then i close the window and come back later. But here is my response, Yes it was hitting the full 8GB and i will test again later on today. Video stuttering topic is now a lost cause to me because if i cannot fully explain what i am experiencing/seeing with it, then forget i ever asked.
3.I will take this as opinion. And i also failed to respond to your BF multiplayer question: Yes the results were similar. About 2-5Fps difference. Same map on same 60hz server. This also goes for the other MP games i tested. similar results.
4. Bioshock infinite and sleeping dogs are the only games i know of that have integrated benchmarks in my games list. the only reason why i did not post sleeping dogs Integrated benchmark was because i found out about it after i got to the 7700k when i was checking settings it was right on the bottom of the screen and i missed it. But i went ahead posting my SP run through thinking that it would show a valid result which to me, it did after 2 of the runs i did. I know that Counter Strike Global offensive also has a benchmark tool available on steam workshop but i could not figure out how to run it. I am sure that i posted somewhere stating this but either i have a poor attention span or no one actually responded to it.
The minimum and maximum is good data for me. The maximum is only differs 0.6%. The average differs 4.8%. My outlook is the difference in CPU perfomance from single player to multiplayer with the same game also if more cores help in multiplayer and it looks like the stats are close enough to determine that.
Thanks, that looks good.
Last edited:
1. They do not have integrated benchmarks....not all do. We run BF4 SP, which also doesn't have a benchmark, but it has a section in one of the missions that is "on rails" which makes it repeatable. We also run Crysis 4 in SP manual as well. We have a specific start and stop point and run through killing the same enemys each time through the same path minimizing (but not eliminating) run differences. SP run throughs are just not optimal, but they aren't junk like MP can be
. Games like Civ 6, GRID:Autosport, Shadow of Mordor, Ashes of The Singularity, Rise of the Tomb Raider, Theif, Total War: Warhammer, etc have them among others. 2. Cool, yeah, hitting the RAM and paging out can be an issue with both gameplay and FPS. Those results which are doing so may be off.
3. You can take it as an opinion all you want. If you don't believe me, believe the reputable website that, for the most part, do not run MP titles for the reasons I stated. If it wasn't a valid concern, they would be used a lot more than they are.
You say 2-5 FPS, but what does that REALLY mean? You need to look at percentages. 2-5 FPS in your BEST case scenario (with the highest FPS) in BF4 is 1-2.5% run variance. In worst case, its ~2-10% run variance. FPS doesn't mean anything when there isn't a scale to compare it to.
4. If Sleeping Dogs has one, use it. The results will be a bit more consistent there in doing so.
Last edited:
OP
- Joined
- Apr 14, 2011
- Thread Starter
- #112
this was after three runs. Ultra settings in game and i selected user graphics settings and user resolution and thats what i got.
like which one would i record?
Alaric
New Member
- Joined
- Dec 4, 2011
- Location
- Satan's Colon, US
Average the three?
OP
- Joined
- Apr 14, 2011
- Thread Starter
- #115
Ok so now i think i understand what im supposed to do.
Pick a part of the game that is easy to repeat over and over. Once found, record results a few times and then record the lowest result you get.
Right? RIGHT?!?!
Pick a part of the game that is easy to repeat over and over. Once found, record results a few times and then record the lowest result you get.
Right? RIGHT?!?!
there will be on manual run throughs, sure. Any change in action will be a change in fps. Typically its less. Not knowing wth is tested in bioshock, who knows why there is such a difference. In MP the issue can be exacerbated due to adding another variable (the network/server). Seems to further suppport the underlying message (mp is junk, sp manual runs throughs not junk also not optimal). I wouldnt jump to any conclusions about mp vs sp over one title regardless. A result of 43 fps of well over 15% difference doesnt raise a red flag??? I can easily play any game different and get varying fps...this should all bring things full circle!!I just shows there is the same variation with single player as multiplayer.
Really ginny, your goal on these sp manual run throughs is to make them as repeatable as possible. You can only do so much without using an integrated benchmark. For sp manual run it 3x and average it.
Last edited:
OP
- Joined
- Apr 14, 2011
- Thread Starter
- #117
I did use the integrated bioshock infinite benchmark and recorded results right at the beginning of the loading screen and then again a little bit after the loading screen.
This is starting to drive me crazy a little. For battlefield 1 and 4, the only difference I see is 10fps after doing three runs.
This is starting to drive me crazy a little. For battlefield 1 and 4, the only difference I see is 10fps after doing three runs.
The integrated benchmark gives you FPS data. That's the point of them and you wouldn't use another frame counter in that case my man . I'd also be willing to bet the values will be more consistent using the in-game benchmark. That doesn't make sense an integrated benchmark is all over the place like that, not at all. We used to run that benchmark a generation or so ago in our GPU testing and I can tell you it wasn't remotely that wonky...something is up there for sure.
10 FPS with 50? 100? 150 FPS? Again, context is needed. 10 FPS difference when talking 50-100 FPS means a lot more than 200 FPS... Its better to include percents so people are aware of the weight the FPS value carries. 10 FPS against 162 is a 7% difference. 10 FPS against 195 is a bit over 5% difference. Still a difference like that when compared to less than 1% which is typical of integrated benchmarks, it sticks out like a sore thumb. So the datasets have, to START a 5-7% variance. Its turning something empircal into a weather forecast.
. I'd also be willing to bet the values will be more consistent using the in-game benchmark. That doesn't make sense an integrated benchmark is all over the place like that, not at all. We used to run that benchmark a generation or so ago in our GPU testing and I can tell you it wasn't remotely that wonky...something is up there for sure. 10 FPS with 50? 100? 150 FPS? Again, context is needed. 10 FPS difference when talking 50-100 FPS means a lot more than 200 FPS... Its better to include percents so people are aware of the weight the FPS value carries. 10 FPS against 162 is a 7% difference. 10 FPS against 195 is a bit over 5% difference. Still a difference like that when compared to less than 1% which is typical of integrated benchmarks, it sticks out like a sore thumb. So the datasets have, to START a 5-7% variance. Its turning something empircal into a weather forecast.
Last edited:
I use to do bench marking like Crysis_Benchmark_Tool- and run after run it was close like his multiplayer benchmarks. Recently I have been doing windows 10 memory benchmarks and they are different all the time in retesting. EarthDog I think you are being to hard on his testing. Minimum and Maximum FPS bench testing is semi accurate with sing player and multiplayer.
Perhaps you are right. Go ahead and bench away and take away from these results what you will. 
I don't think memory testing is any barometer to go by for consistency though.
Min/Max being semi-accurate really doesn't tell much either. That is literally like 1 frame from each in a pool of tens of thousands. Literally the ends of the bell curve.
Anyway, get to it and enjoy Gin.
I don't think memory testing is any barometer to go by for consistency though.
Min/Max being semi-accurate really doesn't tell much either. That is literally like 1 frame from each in a pool of tens of thousands. Literally the ends of the bell curve.
Anyway, get to it and enjoy Gin.
