First results from the 4.11.0 survey

[inkoalawetrust]

Hmm, yeah, that seems to be what this does, so I guess it just doesn't bother dumping the CPU info of generic-ish low to mid end Intel CPUs, instead of being a list of specific GPU vendors that should be sent to the survey. So I take it back.

[Professor Hastig]

From what I'm able to tell (and I am not a programmer by trade), the following block of code:

Code: Select all

		if (device.Compare("AMD Radeon(TM) Graphics") == 0 ||
		device.Compare("Intel(R) UHD Graphics") == 0 ||
		//device.Compare("Intel(R) HD Graphics") == 0 || these are not that interesting so leave them alone
		device.Compare("Intel(R) Iris(R) Plus Graphics") == 0 ||
		device.Compare("Intel(R) Iris(R) Xe Graphics") == 0 ||
		device.Compare("Radeon RX Vega") == 0 ||
		device.Compare("AMD Radeon Series") == 0)
	{
		// for these anonymous series names add the CPU name to get an idea what GPU we really have
		auto ci = DumpCPUInfo(&CPU, true);
		device.AppendFormat(" * %s", ci.GetChars());
	}

is checking the name of the GPU, and if it matches one of the above strings, it takes the extra step to DumpCPUInfo and then format the resulting extra data. That appears to be all that this if block does.

What this block does is quite obvious, it looks for some generic GPU names that span entire multiple generations and adds the CPU name to it so that they become distinguishable. The string "AMD Radeon Series", for example, is used on a large group of iGPUs from a wide ranging list of CPUs - some low end garbage, the other quite nice upper mid range systems.
About that comment for Intel HD Graphics, it's also very simple: Intel's older offerings are easy to distinguish by the maximum supported OpenGL version, so the info probably turned out to be somewhat unnecessary. All iGPUs marked with this string have always been the lowest performers in their respective hardware generation, so by default they all can be classified as 'low performance non-Vulkan supporting hardware' without going into details.


Not everything is a conspiracy, for god's sake!

[yum13241]

Because not "everybody on a PC from before 2016" is getting shafted.
For discrete graphics cards, Vulkan had been supported since 2013, and if you got a laptop with a better GPU the same would apply.
Also, as you can see, it's 8.5% of users who would be forced to use the fallback OpenGL build that won't see any new features in the future, and that number will only shrink further.
Why this survey? Simply because I wanted to see that number as it gives a good idea how long it will take until OpenGL becomes irrelevant. My current guess is 4 years.

And OpenGLES?

[Graf Zahl]

In this context both are the same.

[Graf Zahl]

Regarding the bindless texture issue, the current numbers are:

11% of our user base runs the Vulkan backend. Of these 11% 90% report Vulkan 1.3, a further 9% run Vulkan 1.2, and 15 users overall run Vulkan 1.1 or 1.0.
Of these 15 users 8 are on NVidia, 6 on AMD and 1 on Intel - this one user has a HD630. I can see HD610 and HD620 with Vulkan 1.3 drivers so the HD630 will likely have support as well.

From the survey numbers Vulkan 1.0 and 1.1 look to be completely irrelevant, it's all just old drivers of hardware than can do better.

[Phredreeke]

25% of userbase has RTX cards, but only 11% uses Vulkan? How does that even make sense?

[Graf Zahl]

OpenGL was the default backend,they never changed the setting.

[Blzut3]

Despite Cacodemon345's claims, as far as I can tell the extension in question is support by AMD back to GCN 1 (which doesn't have Vulkan 1.2 but based on mesamatrix.net that's only missing VK_EXT_scalar_block_layout, so if that's to be believed it's basically Vulkan 1.3). So of the three PC GPU vendors this would only affect those on Linux using the hasvk (Ivy Bridge, Haswell, and Broadwell) driver since those never had a Windows driver. Otherwise the only reason to not have it would seem to be running out of date drivers.

[Professor Hastig]

AMD did have bindless textures on that generation for OpenGL so unless they did the Intel thing and stopped updating drivers it should support it.

[Graf Zahl]

Those generations of Intel chipsets are currently 0.6% of the total user base and 6% of the Linux user base, altogether some 50 out of 9000 reports. That shouldn't hold us back, GLES would be acceptable for those.

[yum13241]

GLES would be acceptable for those.

I don't know why, but the latest GZDoom on GLES runs faster than the latest LZDoom on GL. It seems that GL is really is on it's way out, heh. Either way, it's more than acceptable for me.

[Graf Zahl]

That's really no surprise. The GLES backend has some significant shader performance optimizations which LZDoom does not have. Our own tests were not much different on most cards, at worst they were about equal.

[Graf Zahl]

Yesterday we got a survey report that really baffles me.
There's someone claiming to run a Geforce 3050 on Windows Vista!
Is this even possible?

It's also the only report claiming to run on Vista.

[dpJudas]

With some of the hacks I've witnessed in the UT99 community, if a driver can be hacked to boot up on Vista, and if that driver is compatible enough with the hardware and/or driver architecture of Vista/7, then yes, someone might be insane enough to do it. I even saw one try humor me by creating a bug report for crashes with this type of setup.

[yum13241]

Yesterday we got a survey report that really baffles me.
There's someone claiming to run a Geforce 3050 on Windows Vista!
Is this even possible?

It's also the only report claiming to run on Vista.

Someone (Enderman) got Windows 11 to run on a Pentium 4. You know, the same Enderman who made destroyed Windows 10 with ads.