GPT4All v2.7.1 appears in taskbar but does not show a window

[jonarser]

I use Windows 11. Intel GT710M graphics card (but I only use CPU), Intel Core i3x processor. After updating the program to version 2.7.1, it stopped running. Only the icon on the taskbar appears, and the application takes up 4 gigabytes of RAM. The window does not open, even after a ten-minute wait. I installed version 2.6.x, which turned out to be working.
Even for some people with powerful computers, the program did not run. Is this somehow related to the absence of AVX, then can I leave this as an optional option?

[cebtenzzre]

Intel Core i3x processor

What is the exact model of your CPU? Go to Settings > System > About. Only the first generation of i3 CPUs (Clarkdale, i3-5XX) should be unsupported. Sandy Bridge (i3-2XXX) and above should work.

[jonarser]

Intel Core i3x processor

What is the exact model of your CPU? Go to Settings > System > About. Only the first generation of i3 CPUs (Clarkdale, i3-5XX) should be unsupported. Sandy Bridge (i3-2XXX) and above should work.

I have i3-3217U Ivy Bridge
It's sad that not all architectures are supported now. Why is this so?
Will future versions support my architecture?

[manyoso]

I have i3-3217U Ivy Bridge It's sad that not all architectures are supported now. Why is this so? Will future versions support my architecture?

To put it simply the CPU that you have just doesn't have the modern x86 instructions necessary to run these models in a way that isn't super slow. It is a hardware limitation and the experience you'd have running models on such an old CPU does not warrant the amount of work required to support it and maintain it. I'm sorry, but this CPU is just too old to make it work well.

[jonarser]

I have i3-3217U Ivy Bridge It's sad that not all architectures are supported now. Why is this so? Will future versions support my architecture?

To put it simply the CPU that you have just doesn't have the modern x86 instructions necessary to run these models in a way that isn't super slow. It is a hardware limitation and the experience you'd have running models on such an old CPU does not warrant the amount of work required to support it and maintain it. I'm sorry, but this CPU is just too old to make it work well.

But I have AVX

[cebtenzzre]

I'm sorry, but this CPU is just too old to make it work well.

This issue does not seem to be CPU-related, because GPT4All still runs fine on my Sandy Bridge Xeon E5-2650L machine, which I use to test AMD GPU support. Ivy Bridge is a generation newer, and should have no issue. There's something else going on here.

[cebtenzzre]

@jonarser If you'd like to join the Discord and ping me (@cebtenzzre), I can help you troubleshoot this further. I would love to get a stack trace from procdump on a debug build of v2.7.3.

[cebtenzzre]

The call stacks don't show anything interesting - all threads are just blocked in WaitForMultipleObjects or similar.

[jonarser]

Can I record other dumps? Or could the error be related to Cyrillic characters in the path (it is set automatically) where the cache is written? The program itself is installed along the path with only Latin characters. I will try to install the regular version on another account.

[jonarser]

My hypothesis was confirmed. Apparently, the error occurs when working with the cache. You can select a path consisting only of the Latin alphabet only during installation. And the cache path cannot be selected, and it is written somewhere in the user's folder, and it contains a non-Latin character. Everything worked with a different account. Maybe during installation it offers to select the cache path, or configure work with any characters in UTF-8 format?

[cebtenzzre]

This should be fixed in GPT4All version 2.7.4.

[jonarser]

thx