gerwin wrote on 2024-01-26, 19:32:

I agree about the Atom situation, as it was back then. It was a nice idea, but they never got it right in the Windows XP driver […]
Show full quote

386SX wrote on 2024-01-26, 19:07:

Imho the problem for Intel was most probably the lack of a time updated own GPU which let the early Atoms totally alone without a better GPU acceleration even for the new 2D GUIs before 3D gaming. Also if I remember correctly some SGX version didn't even have the old style 2D acceleration but more oriented to OpenGL ES smartphone UI. But the later Atom cut power demands quite a lot while the old GMA + southbridge platforms required much more power than the CPU itself, it was really pointless to put that in any mobile devices.

The E-350 went too far with its Radeon iGPU and with a very weak CPU anyway that is strangely too slow, maybe for the lack of SSE4 instructions but as an out of order architecture some was expecting more. The fan noise in my brand new ITX industrial config is something impossible to even accept. It was too slow for any notebook like platforms and too power demanding for a netbook/smartphone like device.

I agree about the Atom situation, as it was back then. It was a nice idea, but they never got it right in the Windows XP driver era. After that I lost interest. Though in some limited use-cases these Atoms were fine anyways.

As for the E-350. The CPU is weak compared to the Graphics indeed. But it is a nice hobby box for some open-source game projects, which expect a certain OpenGL-version and the power to actually use it. I also found it working well as a SNES emulator. Besides, connecting it to a TV with 1920x1080 screen already puts some demands on the GPU.
That fan noise, would it be fair to hold against the E-350 in general? It is actually the system designer at fault. Surely a system can be designed to dissipate the heat reasonably. Or maybe it is because of BIOS and/or OS power-saving support?
Passive PCIe cards of the similar Radeon HD 6450 were sold. I must admit that these could get too hot when used in ITX-sized PC cases. ( NVidia GT 710 remains cooler, and is faster. GT 710 is three years more modern, but still supports Windows XP as well. )

Of course the E-350 wasn't really bad just went too far for what it should have been sold for as an Atom alternative. If Intel didn't have the right GPU, AMD probably didn't have the right CPU.

Minutemanqvs wrote on 2024-01-25, 17:56:

What about a HP 2133 for a small C7 based system?

https://en.wikipedia.org/wiki/HP_2133_Mini-Note_PC

I have one. Despite it is very good in theory, in practice it isn't.
The main problem is that despite the stated compatibility of VIA Chrome9HC with Windows 9X in the drivers of later VIA chips, such as the CN896 or VX700, in practice these drivers do not work in Windows 9x anywhere, perhaps except on VIA motherboards itself.
I assume that this problem can be solved by modifying BIOSes or drivers, but I do not have sufficient skills to do this. Perhaps some of the enthusiasts could.
Second problem is an integrated VIA HD Audio which isn't supported in Windows 9x.

I experimented a bit with thin clients to check wether they are suitable for compact Win 98 machines.
@Kruton is correct when he states that the later VIA chipsets for the C7 are not really suited for Windows 98. The graphics driver is more trouble than anything and there are no drivers for the HD sound.
Most thin clients come without a PCI slot and some have just a single one. So you either adress the graphics problem with a PCI card and tackle the sound via USB sound card or you install a sound card like the Solo-1 and try to live with the problem-ladden internal graphic.
If you can settle for lower horse power, you can shoot for an older samuel II 533 or similar and an older chipset like PLE133, which has Soundblaster compatible audio hardware and a Trident Blade3D-type hardware integrated.
Speed-wise you are then in Pentium 233MMX territory, but with a smaller footprint. Of course, your upgrade options are then severily limited.

BitWrangler wrote on 2024-01-24, 16:33:

Hi folks, […]
Show full quote

Hi folks,

So I am looking at some atom/gma500 platforms as potentially fun to mess with for retro. Problem is, it was really the Vista era when they were around, but in terms of gaming horsepower they're more like late Win98.

So they'd be stuck with the built in GMA500 graphics which is supposed to have PowerVR IP core inside, but I gather that none of the earlier discrete PowerVR stuff works. But for the GMA 500 itself has anyone heard of ME or 98 compatible drivers? I believe XP drivers are around and can be massaged to work in 2000.

Also wondering about DOS screen modes and if there's any glaring omissions for using ~92 up DOS games.

Thanks for any clues...

Just curious, did you ever get around to messing with any GMA500 (PowerVR) equipped Atom systems?

I was reading about Kyro cards and it got me wondering if PowerVR's stealthy and unassuming return to the PC market via GMA500 was even remotely worth messing with. Like, are the drivers decent enough to actually play 3D games? How far back does compatibility go?

I found this whole website dedicated to keeping this platform alive, at least for a while. https://gma500booster.blogspot.com/2017/09/download.html

In my imagination (hah, get it?), I picture one of these Atom systems with a PowerVR graphics chip from ~2008-2010 being at least as good for early 2000s gaming as an Athlon or PIII with a Kyro 2 from 2001... but those were very different times. I can also imagine the SGX535 making it all the way to retail x86 products without anyone on the driver team even considering game compatibility under Windows XP or Vista.

It's surprising how little info there is about people actually using 2008-2010 netbook-level hardware for software from 8-10 years prior. To me, that seems like the best possible use for them since they were BARELY adequate for contemporary computing when they were released.

afaik the drivers are a complete disaster on this Intel implementation of the powervr design, it's practically useless and performs very poorly on anything, the GMA 950 and its variants is far better
it's very difficult to find info, but it's for a good reason
there are some performance numbers for the top version (GMA 3650) in here
https://www.notebookcheck.info/Intel-Graphics … 50.90432.0.html

on youtube there are some videos, but it's always running games at 15fps
https://www.youtube.com/playlist?list=PLv7kwt … U9UsVdqgY-38RRy

quake 3 seems kinda ok, UT is really bad, gta 3 really bad, and newer games...

SPBHM wrote on 2025-02-14, 18:06:

afaik the drivers are a complete disaster on this Intel implementation of the powervr design, it's practically useless and perfo […]
Show full quote

afaik the drivers are a complete disaster on this Intel implementation of the powervr design, it's practically useless and performs very poorly on anything, the GMA 950 and its variants is far better
it's very difficult to find info, but it's for a good reason
there are some performance numbers for the top version (GMA 3650) in here
https://www.notebookcheck.info/Intel-Graphics … 50.90432.0.html

on youtube there are some videos, but it's always running games at 15fps
https://www.youtube.com/playlist?list=PLv7kwt … U9UsVdqgY-38RRy

quake 3 seems kinda ok, UT is really bad, gta 3 really bad, and newer games...

Oh wow, that is horrifically bad. What a shame. I'm sure the hardware was capable of doing a bit better, but that is just awful.

I remember reading about the GMA 3600\3650 a while back when the Atom N2xxx series started showing up and on paper it looked pretty decent. But any time I saw it benchmarked it was... at the bottom.

Too bad the implementation was so bad. PowerVR's chips were extremely successful in non-Windows mobile devices.

Ozzuneoj wrote on 2025-02-14, 17:37:

Just curious, did you ever get around to messing with any GMA500 (PowerVR) equipped Atom systems? […]
Show full quote

BitWrangler wrote on 2024-01-24, 16:33:

Hi folks, […]
Show full quote

Hi folks,

So I am looking at some atom/gma500 platforms as potentially fun to mess with for retro. Problem is, it was really the Vista era when they were around, but in terms of gaming horsepower they're more like late Win98.

So they'd be stuck with the built in GMA500 graphics which is supposed to have PowerVR IP core inside, but I gather that none of the earlier discrete PowerVR stuff works. But for the GMA 500 itself has anyone heard of ME or 98 compatible drivers? I believe XP drivers are around and can be massaged to work in 2000.

Also wondering about DOS screen modes and if there's any glaring omissions for using ~92 up DOS games.

Thanks for any clues...

Just curious, did you ever get around to messing with any GMA500 (PowerVR) equipped Atom systems?

I was reading about Kyro cards and it got me wondering if PowerVR's stealthy and unassuming return to the PC market via GMA500 was even remotely worth messing with. Like, are the drivers decent enough to actually play 3D games? How far back does compatibility go?

I found this whole website dedicated to keeping this platform alive, at least for a while. https://gma500booster.blogspot.com/2017/09/download.html

In my imagination (hah, get it?), I picture one of these Atom systems with a PowerVR graphics chip from ~2008-2010 being at least as good for early 2000s gaming as an Athlon or PIII with a Kyro 2 from 2001... but those were very different times. I can also imagine the SGX535 making it all the way to retail x86 products without anyone on the driver team even considering game compatibility under Windows XP or Vista.

It's surprising how little info there is about people actually using 2008-2010 netbook-level hardware for software from 8-10 years prior. To me, that seems like the best possible use for them since they were BARELY adequate for contemporary computing when they were released.

No I didn't get the hardware in hand. At the time there were several handheld machines around cheap, and I was curious of possibilities, but they were looking slim to none. Possibly I might get to try some stuff on GMA 3600 in future as that chipset is in the house in a netbook of my wife's which she may give up on now support for highest version of win7 compatible browsers is expiring from many websites. However, it has also got really terrible support. IDK if stuff would have to be backported from ARM Android/Linux to get anything like it's potential unleashed.

Imho the problem with these SoC I tested for so long in the last decade until I had no mainboard working was that the core inside these platforms wasn't really oriented to desktop/Win PCs but old Android/Linux/OpenGL-ES configs simpler o.s. It was working and probably could have worked better if more supported by the companies but then again the choosen solution inside those Atom wasn't quite simply powerful enough to compete with most standard GPUs cause it would have been needed very specific game programming/design/optimizations to really perform and it was really a smartphone oriented GPU installed in a Win oriented SoC for power/costs reasons. The choosen config inside the "fast" SGX545 was one of the slowest probably for power demand limits, limited core area at that size, limits of the Atom platforms, limits of the netbook project. But power demand was probably the main reason.

The PowerVR IC could have been scalable @ many more unified engines but would have required MUCH more power. Also the complexity of the driver design felt quite heavy with binaries, micro kernel, etc like a lot of work for not a real motivation at that point when netbooks were not gaming machines and Win 8.x changes/complexity was added. I liked so much these solutions cause they felt really alternative and they really were actually. But for real life usage not to mention retrogaming or modern gaming was a pain. Drivers developments felt very difficult and with not great changes and at best they felt like the ATi Rage 2C driver evolution with much CPU usage and really strange FPS results. On the good side the SGX config into these Atom were almost power free, the decoding media engine for H264 1080p even 60fps was really good when running on Windows players.

EDIT: I've seen after writing that I already wrote in this old thread but anyway still interesting subject.

I would add a very important aspect to this discussion. Even if some graphics chip has win9x drivers, that doesn't means the actual 9x era video games will work properly. Early 9x games could have serious problems even on newer cards than a tnt2.

For win9x gaming, laptops are generally a bad choice, because when laptops finally started to have usable 3d cards, that was well into the win XP era. A typical 300 mhz laptop will usually have some crappy 2 mbyte integrated svga chip. A p4 laptop in other hand will have a radeon. And its hard to find anything in between.

Also these SGX solution had very low multi texturing fill rate numbers. But imho Win 9x style o.s. would have been mostly not possible, the SGX545 was a unified proprietary architecture similar to a Directx 10 iGPU. I had to tweak a lot Directx6 games to run and finish them and low fps, in specific Thief 1 and Thief 2 games with all sort of tricks to get some 20/25fps@1024x768 with Directx6 native calls and a bit better with Directx9 modern path but with other problems to solve too in the rendering of the old 2D menus.

I should see if dgvoodoo works on it for lolz and trollz

BitWrangler wrote on 2025-02-15, 02:00:

I should see if dgvoodoo works on it for lolz and trollz

Oh man! Do it! What if it turned out it that it just "worked" for that purpose and was suddenly fast enough to handle late 90s games with ease... 🤣

386SX wrote on 2025-02-14, 22:53:

Imho the problem with these SoC I tested for so long in the last decade until I had no mainboard working was that the core inside these platforms wasn't really oriented to desktop/Win PCs but old Android/Linux/OpenGL-ES configs simpler o.s. It was working and probably could have worked better if more supported by the companies but then again the choosen solution inside those Atom wasn't quite simply powerful enough to compete with most standard GPUs cause it would have been needed very specific game programming/design/optimizations to really perform and it was really a smartphone oriented GPU installed in a Win oriented SoC for power/costs reasons. The choosen config inside the "fast" SGX545 was one of the slowest probably for power demand limits, limited core area at that size, limits of the Atom platforms, limits of the netbook project. But power demand was probably the main reason.

The PowerVR IC could have been scalable @ many more unified engines but would have required MUCH more power. Also the complexity of the driver design felt quite heavy with binaries, micro kernel, etc like a lot of work for not a real motivation at that point when netbooks were not gaming machines and Win 8.x changes/complexity was added. I liked so much these solutions cause they felt really alternative and they really were actually. But for real life usage not to mention retrogaming or modern gaming was a pain. Drivers developments felt very difficult and with not great changes and at best they felt like the ATi Rage 2C driver evolution with much CPU usage and really strange FPS results. On the good side the SGX config into these Atom were almost power free, the decoding media engine for H264 1080p even 60fps was really good when running on Windows players.

EDIT: I've seen after writing that I already wrote in this old thread but anyway still interesting subject.

Yeah, it seems they really used an underpowered variant of the SGX, which just made it all the worse when the software wasn't carefully tuned to make the best use of it in Windows apps\games. The model in the 2012 iPad 4 (SGX 554MP4) was 3-8x as fast in some cross platform benchmarks according to this page (vs the VivoTab Smart with an Atom Z2560 + PowerVR SGX 545):
https://www.anandtech.com/show/6872/the-great … ldxbenchmark-27

Ozzuneoj wrote on 2025-02-15, 02:07:

Oh man! Do it! What if it turned out it that it just "worked" for that purpose and was suddenly fast enough to handle late 90s g […]
Show full quote

BitWrangler wrote on 2025-02-15, 02:00:

I should see if dgvoodoo works on it for lolz and trollz

Oh man! Do it! What if it turned out it that it just "worked" for that purpose and was suddenly fast enough to handle late 90s games with ease... 🤣

386SX wrote on 2025-02-14, 22:53:

Imho the problem with these SoC I tested for so long in the last decade until I had no mainboard working was that the core inside these platforms wasn't really oriented to desktop/Win PCs but old Android/Linux/OpenGL-ES configs simpler o.s. It was working and probably could have worked better if more supported by the companies but then again the choosen solution inside those Atom wasn't quite simply powerful enough to compete with most standard GPUs cause it would have been needed very specific game programming/design/optimizations to really perform and it was really a smartphone oriented GPU installed in a Win oriented SoC for power/costs reasons. The choosen config inside the "fast" SGX545 was one of the slowest probably for power demand limits, limited core area at that size, limits of the Atom platforms, limits of the netbook project. But power demand was probably the main reason.

The PowerVR IC could have been scalable @ many more unified engines but would have required MUCH more power. Also the complexity of the driver design felt quite heavy with binaries, micro kernel, etc like a lot of work for not a real motivation at that point when netbooks were not gaming machines and Win 8.x changes/complexity was added. I liked so much these solutions cause they felt really alternative and they really were actually. But for real life usage not to mention retrogaming or modern gaming was a pain. Drivers developments felt very difficult and with not great changes and at best they felt like the ATi Rage 2C driver evolution with much CPU usage and really strange FPS results. On the good side the SGX config into these Atom were almost power free, the decoding media engine for H264 1080p even 60fps was really good when running on Windows players.

EDIT: I've seen after writing that I already wrote in this old thread but anyway still interesting subject.

Yeah, it seems they really used an underpowered variant of the SGX, which just made it all the worse when the software wasn't carefully tuned to make the best use of it in Windows apps\games. The model in the 2012 iPad 4 (SGX 554MP4) was 3-8x as fast in some cross platform benchmarks according to this page (vs the VivoTab Smart with an Atom Z2560 + PowerVR SGX 545):
https://www.anandtech.com/show/6872/the-great … ldxbenchmark-27

Yes, all the test I've done with old games and "modern" Directx9 games/bench showed the limit of that solution and the drivers complexity was indeed there with a very alternative and not-standard GPU design. Maybe for compatibility, maybe for a cross platforms adaptation reasons. Modern features like Pixel Shading support was good and quite fast when considering the few watts required but GUI acceleration ended up feeling slower than the CPU sw accelerated version.
Also the SGX545 had 400Mhz core clocks while the 3650 pushed it to 640Mhz if I remember correctly, still not enough on a single channel DDR3 low power shared memory. The whole point of the SGX solution was to have a functional modern iGPU cheap and technically capable beside it ended up showing why desktop GPUs were still needed in a desktop oriented machine. Still nowdays with very powerful Raspberry Pi SBC I'd say they are far from any low end GFX intensive o.s. scenario.

BitWrangler wrote on 2025-02-15, 02:00:

I should see if dgvoodoo works on it for lolz and trollz

If I remember I tried all sort of wrapper in these solutions and ended up introducing even more overhead/latency without real possible benefits. While the slow CPU wasn't always the real problem here, it was indeed slow and the memory performances too. What it was needed here was the opposite, lower lever optimizations at driver levels and more standard driver architecture if even possible. Anyway when considered using Windows 7 Starter o.s. only, the experience of these iGPU are mostly acceptable with the correct WDDM compatibility required by the Win 7 only working driver. The Linux situation instead for years has been an horror movie that might be an idea for a whole TV serie about it..

In this thread there are some tests that might be of interest, in case you haven't seen it already.
Gaming on my Intel Atom

https://www.youtube.com/watch?v=CJjTRJKllt8

yeah...i have that netbook among others, the GMA 500 kills all the fun. There's only drivers for windows vista and 7

i think its one of the worst IGPUs i ever seen