How can I see if Mint 19.3 and 20.2 use GPU part from this APU?

glxgears works with 60 fps

On 11/6/2021 12:14 AM, Mrde wrote:
> How can I see if Mint 19.3 and 20.2 use GPU part from this APU?
>
> glxgears works with 60 fps

Here is mine.

[Picture]

https://i.postimg.cc/V6y5G3zW/glxgear-vblank-zero.gif

vblank_mode=0 glxgears

Things typed before the command invocation, are
passed as part of the environment. Maybe usage of "export"
could make such a setting more permanent (until the end
of the session).

If you used such a setting before playing a 3D game,
you might see image tearing, so the setting is
not always the best choice. But for doing burnouts
on a PC, it's a perfect choice :-)

Happy benching,
Paul

On 11/6/2021 2:31 AM, Paul wrote:
> On 11/6/2021 12:14 AM, Mrde wrote:
>> How can I see if Mint 19.3 and 20.2 use GPU part from this APU?
>>
>> glxgears works with 60 fps
>
> Here is mine.
>
> [Picture]
>
> https://i.postimg.cc/V6y5G3zW/glxgear-vblank-zero.gif
>
>    vblank_mode=0 glxgears
>
> Things typed before the command invocation, are
> passed as part of the environment. Maybe usage of "export"
> could make such a setting more permanent (until the end
> of the session).
>
> If you used such a setting before playing a 3D game,
> you might see image tearing, so the setting is
> not always the best choice. But for doing burnouts
> on a PC, it's a perfect choice :-)
>
> Happy benching,
>    Paul

Oh, one other thing. I don't think this method is
necessary "linear". Interrupt rate handling may be
preventing the figure from going even higher. So
the owner of a Titan, may not get the 100,000
they deserve. Maybe all the hot graphics devices
stop at 25000. You can compare your new CPU and its GPU,
to some other cards, and see if there is a
suspicious upper limit. I can't remember the results
I was getting, but I had some results that did not
make sense particularly (two cards that were too close
to one another).

AMD made their own version of GLXGears, which was a spinning
oscillating figure in 3D. But I don't think that is available
separately.

You can run the application "radeontop", which gives stats
on GPU usage. As it turns out, that little graphic is using
most of my Vega 7 puny GPU :-) On NVidia, they have
"nvidia-smi" for indicating usage, in a terminal text window,
which is applicable if you have the NVidia driver installed.

Paul

On 11/6/21 7:31 AM, Paul wrote:
> On 11/6/2021 12:14 AM, Mrde wrote:
>> How can I see if Mint 19.3 and 20.2 use GPU part from this APU?
>>
>> glxgears works with 60 fps
>
> Here is mine.
>
> [Picture]
>
> https://i.postimg.cc/V6y5G3zW/glxgear-vblank-zero.gif
>
>    vblank_mode=0 glxgears

AMD 5 3500U with vblank have around 5200 fps

Is it normal for this APU?

On 11/7/2021 8:52 PM, Mrde wrote:
> On 11/6/21 7:31 AM, Paul wrote:
>> On 11/6/2021 12:14 AM, Mrde wrote:
>>> How can I see if Mint 19.3 and 20.2 use GPU part from this APU?
>>>
>>> glxgears works with 60 fps
>>
>> Here is mine.
>>
>> [Picture]
>>
>> https://i.postimg.cc/V6y5G3zW/glxgear-vblank-zero.gif
>>
>>     vblank_mode=0 glxgears
>
> AMD 5 3500U with vblank have around 5200 fps
>
> Is it normal for this APU?

My CPU is 65W power envelope, with desktop cooler. It
never seems to get too hot, and it has on occasion drawn
65W approximately (checked with clamp-on DC meter rather
than Kill-O-Watt meter). The entry for yours says it is
normally set for TDP 15W, while 35W TDP is possible if
it has a regular laptop cooler say. The 15W TDP would be
devices like tablets, where the skin does the cooling,
and 15W is probably a stretch for skin cooling (uncomfortable).

This is Passmark G3D and G2D benchmark numbers. Unfortunately,
the Vega8 is not available as a APU entry in the web page, so
there is some uncertain what the second entry is measuring. It
appears to be a bit lower, but not in the ratio of 23500 to 5200.
At a guess, something else (throttling) is happening.

Ryzen 5 5600G with Radeon Vega7 Graphics 2,638 829 NA Unknown

https://www.videocardbenchmark.net/gpu.php?gpu=Radeon+Vega+8&id=3895

Vega8 Average G3D & G2D Mark = 1,597 424 (partially lower, due to clocking,
not because it lacks shaders)

Check to see, with RadeonTop and regular "top", whether
the utilization looks reasonable.

[Picture]

https://i.postimg.cc/DyvYX5Dj/glxgears-activity-check.gif

Mine never seems to turbo to exactly the book value,
and I don't know if that is normal. Mine has
DDR4-3200 RAM in it, and on purpose, has as many ranks
as desktop memory can manage. Like, 8GB sticks can be
single-rank today, so four sticks gives four ranks.
You need to buy higher capacity sticks than that, to
get double-sided RAM. My setup should have eight ranks,
and the idea is, since the speed and CAS are pretty
flat these days, the only other variable left, is
to do the best you can on ranks. Open pages in RAM,
respond faster to a subsequent hit on the same page
and that's why you want as many ranks and interleaving
to be working on the CPU. This gives a few percent more
of something (I could see a slight difference).

I would say, at a guess, you're a bit low. Maybe it's
doing the best it can, if thermally throttled. It could
be the 15W limit that does it. Skin cooling versus
fan cooling.

Paul

On 11/8/21 1:56 PM, Paul wrote:
>>
>> AMD 5 3500U with vblank have around 5200 fps
>>
>> Is it normal for this APU?
>
> My CPU is 65W power envelope, with desktop cooler. It
> never seems to get too hot, and it has on occasion drawn
> 65W approximately (checked with clamp-on DC meter rather
> than Kill-O-Watt meter). The entry for yours says it is
> normally set for TDP 15W, while 35W TDP is possible if
> it has a regular laptop cooler say. The 15W TDP would be
> devices like tablets, where the skin does the cooling,
> and 15W is probably a stretch for skin cooling (uncomfortable).
>
> This is Passmark G3D and G2D benchmark numbers. Unfortunately,
> the Vega8 is not available as a APU entry in the web page, so
> there is some uncertain what the second entry is measuring. It
> appears to be a bit lower, but not in the ratio of 23500 to 5200.
> At a guess, something else (throttling) is happening.
>
> Ryzen 5 5600G with Radeon Vega7 Graphics   2,638   829   NA   Unknown
>
> https://www.videocardbenchmark.net/gpu.php?gpu=Radeon+Vega+8&id=3895
>
>         Vega8     Average G3D & G2D Mark = 1,597   424       (partially
> lower, due to clocking,
>                                                               not
> because it lacks shaders)
>
> Check to see, with RadeonTop and regular "top", whether
> the utilization looks reasonable.
>
> [Picture]
>
> https://i.postimg.cc/DyvYX5Dj/glxgears-activity-check.gif
>
> Mine never seems to turbo to exactly the book value,
> and I don't know if that is normal. Mine has
> DDR4-3200 RAM in it, and on purpose, has as many ranks
> as desktop memory can manage. Like, 8GB sticks can be
> single-rank today, so four sticks gives four ranks.
> You need to buy higher capacity sticks than that, to
> get double-sided RAM. My setup should have eight ranks,
> and the idea is, since the speed and CAS are pretty
> flat these days, the only other variable left, is
> to do the best you can on ranks. Open pages in RAM,
> respond faster to a subsequent hit on the same page
> and that's why you want as many ranks and interleaving
> to be working on the CPU. This gives a few percent more
> of something (I could see a slight difference).
>
> I would say, at a guess, you're a bit low. Maybe it's
> doing the best it can, if thermally throttled. It could
> be the 15W limit that does it. Skin cooling versus
> fan cooling.
>
>    Paul
>

Now with new Mint 20 I get more then 8800 fps.
So, it's about old Mint 19.

On 11/8/2021 8:06 PM, Mrde wrote:
> On 11/8/21 1:56 PM, Paul wrote:
>>>
>>> AMD 5 3500U with vblank have around 5200 fps
>>>
>>> Is it normal for this APU?
>>
>> My CPU is 65W power envelope, with desktop cooler. It
>> never seems to get too hot, and it has on occasion drawn
>> 65W approximately (checked with clamp-on DC meter rather
>> than Kill-O-Watt meter). The entry for yours says it is
>> normally set for TDP 15W, while 35W TDP is possible if
>> it has a regular laptop cooler say. The 15W TDP would be
>> devices like tablets, where the skin does the cooling,
>> and 15W is probably a stretch for skin cooling (uncomfortable).
>>
>> This is Passmark G3D and G2D benchmark numbers. Unfortunately,
>> the Vega8 is not available as a APU entry in the web page, so
>> there is some uncertain what the second entry is measuring. It
>> appears to be a bit lower, but not in the ratio of 23500 to 5200.
>> At a guess, something else (throttling) is happening.
>>
>> Ryzen 5 5600G with Radeon Vega7 Graphics   2,638   829   NA   Unknown
>>
>> https://www.videocardbenchmark.net/gpu.php?gpu=Radeon+Vega+8&id=3895
>>
>>          Vega8     Average G3D & G2D Mark = 1,597   424       (partially lower, due to clocking,
>>                                                                not because it lacks shaders)
>>
>> Check to see, with RadeonTop and regular "top", whether
>> the utilization looks reasonable.
>>
>> [Picture]
>>
>> https://i.postimg.cc/DyvYX5Dj/glxgears-activity-check.gif
>>
>> Mine never seems to turbo to exactly the book value,
>> and I don't know if that is normal. Mine has
>> DDR4-3200 RAM in it, and on purpose, has as many ranks
>> as desktop memory can manage. Like, 8GB sticks can be
>> single-rank today, so four sticks gives four ranks.
>> You need to buy higher capacity sticks than that, to
>> get double-sided RAM. My setup should have eight ranks,
>> and the idea is, since the speed and CAS are pretty
>> flat these days, the only other variable left, is
>> to do the best you can on ranks. Open pages in RAM,
>> respond faster to a subsequent hit on the same page
>> and that's why you want as many ranks and interleaving
>> to be working on the CPU. This gives a few percent more
>> of something (I could see a slight difference).
>>
>> I would say, at a guess, you're a bit low. Maybe it's
>> doing the best it can, if thermally throttled. It could
>> be the 15W limit that does it. Skin cooling versus
>> fan cooling.
>>
>>     Paul
>>
>
> Now with new Mint 20 I get more then 8800 fps.
> So, it's about old Mint 19.

It's hard to say why that would happen though.
And what degree-of-freedom changed.

One part I don't understand about APUs, is there
is one power envelope for the APU (the TDP), and
yet the CPU or the GPU could be turboed to
higher clock values. How does the scheduler or
the power management, know which subsystem deserves
the majority of the power ? That, I don't understand.

Paul