3DMark
What is going on with Port Royal?
I was gradually OCing my video card to best my scores which I was doing and then I finally get my top scores and this happens. My card is running stable and no crashes. I have been running constant tests the past hour or more. What inconsistencies is this talking about? I see none visually. This is crazy. I have been unhappy lately with 3DMark. I may have to change my rating.



https://steamuserimages-a.akamaihd.net/ugc/950724431495542927/B0345D02D8F14F27026AECF337E42B2C010D6272/

https://steamuserimages-a.akamaihd.net/ugc/950724431495547208/3FE74CD51B77A2FC6946C450CD46E822A79CAF6E/

P.S. I DO NOT USE ANY MODS OR ANY BIOS HACKS. I am not that smart. I just use basic overclocking tools like Precision and ASUS Tweak.

And I just ran a Time Spy Extreme bench just to see if it also failed and did not.

http://www.3dmark.com/spy/5875461

So....what is happening? I would like to get back to running Port Royal but I can't if it keeps saying invalid.
Last edited by Game Cooker US Rocks VR; Jan 19, 2019 @ 4:49pm
< >
Showing 1-15 of 18 comments
UL_Jarnis  [developer] Jan 19, 2019 @ 9:42pm 
Give links to the Port Royal scores and I'll take a look and give you full details. Always post links not screenshots if you see anything you want to know more about.

This error generally means "there are more than one timer on your system. They are now disagreeing how long the test took to run, so we cannot validate that the score is legit. We do not know if a second was actually a second, so frames per second may or may not be right".

Most common way of getting this is by adjusting bus clock speed outside BIOS. Windows 10/8.1 is not set up to react to that correctly. This is different from how things were back during Windows 7 day.

If in doubt, reset all settings to default and re-start OCing from a known good state.

Also please do not threaten review bombing because you do not fully understand everything about the benchmark or overclocking. We fully stand behind our product and help if you ask for it; threats are not needed.
Last edited by UL_Jarnis; Jan 19, 2019 @ 9:52pm
Originally posted by UL_Jarnis:
Give links to the Port Royal scores and I'll take a look and give you full details. Always post links not screenshots if you see anything you want to know more about.

This error generally means "there are more than one timer on your system. They are now disagreeing how long the test took to run, so we cannot validate that the score is legit. We do not know if a second was actually a second, so frames per second may or may not be right".

Most common way of getting this is by adjusting bus clock speed outside BIOS. Windows 10/8.1 is not set up to react to that correctly. This is different from how things were back during Windows 7 day.

If in doubt, reset all settings to default and re-start OCing from a known good state.

Also please do not threaten review bombing because you do not fully understand everything about the benchmark or overclocking. We fully stand behind our product and help if you ask for it; threats are not needed.
Why would this happen all of a sudden? Everything was working fine. I clearly said I was doing tests for an hour or more with your program and then all of a sudden it says invalid. I did nothing with Windows or my motherboard settings. I only changed the the OCing of my video card with Precision as I stated. Not sure what you mean by timers. Obviously I did not change that since I am not sure what you mean. And I am not going to reset the Ocing as it took forever to get to where I am stable. Lol. That is like saying reset Windows to fix your problem. My system is fine. I am not that dumb.
I didn't change anything except Precision which a great many people use I am sure. It does nothing but let you overclock the VRAM and GPU with slider bars. It isn't fancy. I wasn't changing a whole bunch of voltages.

Here is two invalid scores:
http://www.3dmark.com/pr/23822
http://www.3dmark.com/pr/23819
Here is a valid score:
http://www.3dmark.com/pr/23798
http://www.3dmark.com/pr/23798
I just ran two more tests back to back. One validated and one did not for the timing thing. Every other benchmark runs fine and does not give me this problem and I have had my system set the way it is for a while. You guys have to fix this detection thing in your app. I have no hacks or anything in my computer. Everything I do is simple overclocking stuff. Very modest. Nothing that should be raising flags in any app. I know enough to know that for a fact.

These were back to back. Meaning right after another. Just now. This also was after installing the latest Nvidia drivers which I did not want to do. Just in case that was the issue.

Valid: http://www.3dmark.com/pr/24142

Invalid: http://www.3dmark.com/pr/24144
UL_Jarnis  [developer] Jan 20, 2019 @ 10:36am 
Your hardware has issues

Real time clock 141.0
Local time 139.722
Tick count 140.484

Up to 1.3 second difference between the timers. Something about your motherboard or the OC settings is messing things up.

And on one of your validated runs

Real time clock 140.0
Local time 139.677
Tick count 140.438

This one is marginally accepted.

On a system without unstable bus clock, those values should not be off by more than 0.1 - 0.2 seconds.

Edit: And to clarify, if you touch bus clock from Windows, you are going to mess up these timers and cause score to be wrong (and invalid). You can do it only from BIOS.
Last edited by UL_Jarnis; Jan 20, 2019 @ 10:48am
Originally posted by UL_Jarnis:
Your hardware has issues

Real time clock 141.0
Local time 139.722
Tick count 140.484

Up to 1.3 second difference between the timers. Something about your motherboard or the OC settings is messing things up.

And on one of your validated runs

Real time clock 140.0
Local time 139.677
Tick count 140.438

This one is marginally accepted.

On a system without unstable bus clock, those values should not be off by more than 0.1 - 0.2 seconds.

Edit: And to clarify, if you touch bus clock from Windows, you are going to mess up these timers and cause score to be wrong (and invalid). You can do it only from BIOS.
As far as I'm concerned my system is stable. I've had no crashes playing any games at 4K or running furmark or any other benchmarking program. And I just basically overclocked my processor and RAM for my motherboard which is fairly easy since the motherboard has easy overclocking features. I didn't really mess with voltages or anything. And as I said before the video card it's just overclocked simply with Asus GPU tweak 2. You just move some sliders and that's it. So I don't think the timing really matters very much if my system is never crashing. The whole goal is to try to push higher frame rates so I'm not sure what the point of benchmarking is if you can't do that.
somethingcool Jan 21, 2019 @ 2:50pm 
>since the motherboard has easy overclocking features

That's likely your problem, it might be dynamically changing the bus speed or have chosen unoptimal settings. Try manually setting your overclock in the BIOS using just the CPU multiplier and stock voltages.
UL_Jarnis  [developer] Jan 21, 2019 @ 9:18pm 
Timers matter because if you overclock (especially bus speed) outside BIOS, your system can no longer tell the time reliably.

In a benchmark that measures "frames per second" that is critical.
Originally posted by somethingcool:
>since the motherboard has easy overclocking features

That's likely your problem, it might be dynamically changing the bus speed or have chosen unoptimal settings. Try manually setting your overclock in the BIOS using just the CPU multiplier and stock voltages.
Thank you but I already did that. I am using the most modest of overclocking. I don't play with voltages as I stated already. I am no expert but I do understand the basics.
I have 9900k and 32 Gb 3200 RAM with RTX 2080ti.
My point is that my system is running stable. I have no problems with anything else. My computer has been running this way practically since I built it. I had to change the settings back after updating the BIOS a few times but that is it. Maybe they should loosen up the calculation requirements in the app.
3DMark is not the only benchmarking program I use. I use like dozens of them. I guess 3DMark is something special because none of the others fail. I paid money for this so I am not happy. Again this should be about pushing FPS not stability. We are trying to beat a score. Non of the "Kings" are going to run there systems at 6Gz on ice or nitrogen for any length of time therefore stability should be the least issue. If the test finishes then the score should stick. Period. That is how I see it.
Originally posted by UL_Jarnis:
Timers matter because if you overclock (especially bus speed) outside BIOS, your system can no longer tell the time reliably.

In a benchmark that measures "frames per second" that is critical.
I never did that though so I don't know why you keep saying that. I only overclocked the GPU through an app like everyone else does. Everything else is done by the BIOS on the motherboard. AND I NEVER MESS WITH VOLTAGES. That is done by the motherboard.
I give up. 3D Mark passes all other tests but fails in Port Royal. Why does it except timings in all the others but not Port Royal. My systen is stable. I just made sure everything was up to date yet again and checked my MB settings and all is within exceptable parameters according to overclocking forums and sites. No crashes or freezes or failures in any other benchmarks or in anything else except your app. I have to give a negative review now. Sorry. You have not helped me fix the issue I am having.
UL_Jarnis  [developer] Jan 22, 2019 @ 9:22pm 
Port Royal is a shorter test, so it is bit more sensitive.

There is nothing I can do to "fix" your issue. Your motherboard timers are not in sync.

If you reset all overclocking, does the problem go away? If it does, it is related to your overclocking and unfortunately that is something that is always up to the user. Hardware vendors do not guarantee overclocking and we cannot guarantee overclocking.
Originally posted by UL_Jarnis:
Port Royal is a shorter test, so it is bit more sensitive.

There is nothing I can do to "fix" your issue. Your motherboard timers are not in sync.

If you reset all overclocking, does the problem go away? If it does, it is related to your overclocking and unfortunately that is something that is always up to the user. Hardware vendors do not guarantee overclocking and we cannot guarantee overclocking.


How does one fix motherboard timers? Does this have to do with the CPU, RAM and does overclocking the video card effect this?

I am simply changing the CPU multipler to 50(9900k retail boxed version not OEM) and the DDR4 RAM to 3600 from 3200. Most everything is on auto except I enabled all cores open and multi-threading. That is about as basic as one can get. I have an Aorus Master MB which is made for overclocking. How far off do the timers have to be to trash a score? We are talking about what? 1-2 fps maybe. I just can't see how that can kill a score when a system is visually stable. I can't even find a utility to messure what you are talking about. Therefore, I can not investigate how to improve what you are saying except not overclocking. Lol. Again this is an example of your app not being friendly to a lot of basic users. People overclock to beat scores. You are saying don't overclock now....because my app won't let you. Sad.
I will erase my negative review but I am still unhappy. My system is stable. I ran a Sisoftware scan and there was one thing that came up as off and that was dynamic clocking on the CPU which I went in the BIOS and didn't even see checked so that may be done on auto. I don't know. I simplified the Oc'ing more which I stated above. My system says stable on every test I run and visually as well as no crashing or freezing. This timing thing is odd. Your other benchmarks do not fail for it so why this one? Maybe expand that timing thing a bit more.
Or run a mini test before the Port Royal (since that is the only one giving me an issue) to check timings and then make an offset algorithm like noise cancellation to adjust the score a bit. I am not a programmer but that is a suggestion.
UL_Jarnis  [developer] Jan 22, 2019 @ 10:22pm 
If not overclocking = no timer skew and overclocking applies it, I would strongly recommend asking Gigabyte what the heck is their BIOS doing.

Dynamic OC modes where BIOS does whatever it wants to "improve" the OC are easy way to run into this problem. If it Dynamically bumps the bus clock, this definitely can explain it.

Like I said, Port Royal, being shorter than others, is bit more sensitive, but the difference is not major.

Good way to debug:

- Write down all settings you have for your OC
- Run stock run, verify no timer issue
- Add one setting, repeat

Once you find which setting causes timer skew, leave that out and use all others. Then ask Gigabyte why that setting influences windows timers to no longer run in sync with motherboard timers.

Motherboard manufacturers go to great lengths to try to find tiny gains that would make their boards to look "the best" in motherboard reviews. They have cheated in the past and they may cheat again. In this case I would not be surprised if some setting has a baked "tweak" that is minor enough so it is not caught by 3DMark timer check in all old tests, but Port Royal, having released *after* the release of the motherboard, catches their "optimization". And that "optimization" gives them 0.2% higher scores or something - just within margin of error so 3DMark does not flag it, but enough so that when you have 10 boards with same chipset side by side, Gigabyte is "the best" by... 0.2%.

Remember that this issue also means your OS can no longer reliably tell time. Windows 10 can mask this by syncing the clock from the network repeatedly, but it is still a motherboard issue.
Last edited by UL_Jarnis; Jan 22, 2019 @ 10:26pm
Originally posted by UL_Jarnis:
If not overclocking = no timer skew and overclocking applies it, I would strongly recommend asking Gigabyte what the heck is their BIOS doing.

Dynamic OC modes where BIOS does whatever it wants to "improve" the OC are easy way to run into this problem. If it Dynamically bumps the bus clock, this definitely can explain it.

Like I said, Port Royal, being shorter than others, is bit more sensitive, but the difference is not major.

Good way to debug:

- Write down all settings you have for your OC
- Run stock run, verify no timer issue
- Add one setting, repeat

Once you find which setting causes timer skew, leave that out and use all others. Then ask Gigabyte why that setting influences windows timers to no longer run in sync with motherboard timers.

Motherboard manufacturers go to great lengths to try to find tiny gains that would make their boards to look "the best" in motherboard reviews. They have cheated in the past and they may cheat again. In this case I would not be surprised if some setting has a baked "tweak" that is minor enough so it is not caught by 3DMark timer check in all old tests, but Port Royal, having released *after* the release of the motherboard, catches their "optimization". And that "optimization" gives them 0.2% higher scores or something - just within margin of error so 3DMark does not flag it, but enough so that when you have 10 boards with same chipset side by side, Gigabyte is "the best" by... 0.2%.

Remember that this issue also means your OS can no longer reliably tell time. Windows 10 can mask this by syncing the clock from the network repeatedly, but it is still a motherboard issue.
Ok thanks for the advice. I'll take this to heart and delve deeper into this issue. Still bummed I can't get scores for all the tests though.
Just one last note if I didn't say it already. The only time I get this error though is from overclocking the video card not my motherboard. Would that have something to do with power versus voltages or something? Or could the card OC clash with the Motherboard trying to communicate with it at the higher speeds?
Regarding the oc'ing: The app software only has 4 settings I mess with. GPU clock speed, VRAM clock speed and the power bar which has to always go to max in order to get the most power draw. And the voltage slider which is upped by % not actual volts. Usually this can not go above 10% on air.
Thanks again for your time.
< >
Showing 1-15 of 18 comments
Per page: 1530 50

Date Posted: Jan 19, 2019 @ 4:32pm
Posts: 18