New computer

[rpglord]

(03-18-2013, 08:25 AM)NaturalViolence Wrote: Assuming you're not using the vbeam hack that would be correct. The problem is the vbeam hack completely fucks with dolphins normal logic and nobody really understands how it works. It could just be screwing with the readings for all we know. So I remain unconvinced until you can verify this with another method. Which nobody has yet.

So you are agreeing with me now that VPS stands for cpu thread speed ? :p
If its showing false reasing, then why is LLE not stuttering when VPS is at fullspeed ? It clearly is showing correct VPS values

(03-18-2013, 08:25 AM)NaturalViolence Wrote: Try an older build where lock threads to cores still exists (before Billiard decided that useful options are cancer). Turn on lock threads to cores. Repeat the test and screencap task manager.

You mean repeat the test with vbeam trick, right ?
I am currently on intel hd 2000, its so slow I am not sure there is point in testing now, what do you think ?
If not, I will test when I get new card

[NaturalViolence]

rpglord Wrote:So now you are agreeing with me now that VPS stands for cpu thread speed ?

No. It measures exactly what neobrain said is measures in that quote. It is the output of the VID (number of screen refreshes per second).

rpglord Wrote:You mean repeat the test with vbeam trick, right ?

Yes. Whatever you did before, do it again.

rpglord Wrote:I am currently on intel hd 2000, its so slow I am not sure there is point in testing now, what do you think ?
If not, I will test when I get new card

Wait for the new card.

[rpglord]

(03-18-2013, 08:35 AM)NaturalViolence Wrote: No. It measures exactly what neobrain said is measures in that quote. It is the output of the VID (number of screen refreshes per second).

And ibm cpu in real hardware is responsible for that, so whats your point ?
Its for all intents and purposes cpu thread speed.

(03-18-2013, 08:35 AM)NaturalViolence Wrote: Wait for the new card.

Ok

BTW, in case you missed it, I edited previous post to explain why I think vbeam trick is not messing with VPS readings

[NaturalViolence]

rpglord Wrote:And ibm cpu in real hardware is responsible for that, so whats your point ?
Its for all intents and purposes cpu thread speed.

A better definition would be "system speed". It's possible for the emulation of other components such as the dsp or gpu to drag it down. So no I don't agree with you.

rpglord Wrote:If its showing false reasing, then why is LLE not stuttering when VPS is at fullspeed ?

Don't know. It changes the dsp timing logic (well the end result anyways).

roglord Wrote:It clearly is showing correct VPS values

Then prove it. Run the test that I gave you.

[rpglord]

(03-18-2013, 08:54 AM)NaturalViolence Wrote: A better definition would be "system speed". It's possible for the emulation of other components such as the dsp or gpu to drag it down. So no I don't agree with you.

Yeah its posible only if "vbeam trick" is not used.
I suggested to developers to investigate this further, and make it not posible that emulation of other components drag it down, but they didnt care.

And besides, my definition is still correct, even if "vbeam trick" is not used. Just because cpu thread is beeing slowed down, does not mean that VPS still will not be showing its speed

Where is your proof that entire jit/jitil or interpreter engine is NOT beeing slowed down when emulation of other components is slower ?
Because if cpu emulation is in fact beeing slowed down because of other emulation threads, then my definiton stands, VPS will always be showing its speed, "vbeam trick" on or off, doesnt metter.

(03-18-2013, 08:54 AM)NaturalViolence Wrote: Then prove it. Run the test that I gave you.

With intel HD ? I think its doable, you still should be able to see what you want, wont you ?
Its just that gpu thread will be much more bottlenecking

[NaturalViolence]

rpglord Wrote:Just because cpu thread is beeing slowed down, does not mean that VPS still will not be showing its speed

"CPU speed" is a bad definition when doing benchmarks. I explained why. It gives the impression (in your case anyways) that low vps = cpu bottleneck. You can use it if you want but I would prefer it if you didn't.

rpglord Wrote:With intel HD ? I think its doable, you still should be able to see what you want, wont you ?
Its just that gpu thread will be much more bottlenecking

Don't do it with Intel HD 2000. I told you not to because it will guarantee a GPU bottleneck.

[rpglord]

(03-18-2013, 09:13 AM)NaturalViolence Wrote: "CPU speed" is a bad definition when doing benchmarks. I explained why. It gives the impression (in your case anyways) that low vps = cpu bottleneck. You can use it if you want but I would prefer it if you didn't.

You meant cpu thread speed, right ?
VPS cant be lower then FPS.
If its the same, it indicates cpu bottleneck when vbream trick is used, on default settings it can be either thread bottlenecking.

(03-18-2013, 09:13 AM)NaturalViolence Wrote: Don't do it with Intel HD 2000. I told you not to because it will guarantee a GPU bottleneck.

Ok, might get a new card tomorow or in a few days

[rpglord]

Here is screenshoot you wanted, NV.
It clearly shows one thread using cpu core 100%, other doesnt

[NaturalViolence]

It shouldn't look like that. Do you have LLE on thread turned on? If so turn it off. I need to be able to identify which thread is which. Also what graphics card and settings were used?

[rpglord]

(03-19-2013, 07:02 AM)NaturalViolence Wrote: It shouldn't look like that. Do you have LLE on thread turned on? If so turn it off. I need to be able to identify which thread is which. Also what graphics card and settings were used?

Ok, here is the pic without using LLE on thread..causes slowdown in both FPS and VPS.
Not much difference when looking at task manager, except now thread which was mainly using one core 100% but droped to 99% now is constantly at 100%. Other thread is still not using cpu core fully.
I am using 660 gtx, standard dolphin settings except "vbeam trick", and 4xIR but no AA, to make sure not too much of a load is placed on a gpu, and of course "lock threads to cores".
Without it, all four cpu cores are loaded equally, around 50%