ReactOS "Open-Source Windows" Making Progress On SMP/Multi-Core Support

[Rabiator]

My question is, if Reactos had a Windows 98 theme, Windows XP stability and Windows 7 games compatibility, would there be any incentive for Valve to push Linux?

There is also hardware driver compatibility to consider. ReactOS would need to stay up to date with the Windows 10 and 11 driver models so it can use current drivers. If ReactOS only works with Windows 7 drivers, current Nvidia cards won't work according to Wikipedia. AMD has Windows 7 driver support limited to the RX 6600 and RX 6600 XT in the current generation. Personally, I'm surprised that any RX 6000 GPUs still get Windows 7 drivers.

Linux has the advantage that hardware vendors adapt to Linux to some degree. AMD and Intel have open source drivers, it does not get better than that. Nvidia is less forthcoming but still provides a binary driver for Linux.

[hotaru]

Michael, you're off about 5 years on multicore support in PCs. The first multicore CPUs didn't start appearing till around 2007ish.

2001: https://en.wikipedia.org/wiki/POWER4

[Volta]

Your comment reminded me of[...]

I try to make the jump to daily driving Linux on the desktop every year, and so far there have always been deal-breaking issues for me, unfortunately. But my family is quite happy with their Linux web browsing boxes, so I guess there's hope.

I know this one, but thankfully flash is dead.

[Volta]

Reactos would have been ideal for the Steam Deck...

It wouldn't. One broken garbage is enough. Don't let it spread.

[dacha]

Much of their Russian funding has been said to be from Putin, so I wonder if they will be getting even more funding now. I'm sure the Russian military could find uses for ReactOS.

Didn't Putin order the Russian government to use Linux since 2010:

https://linux.slashdot.org/story/10/12/27/2025258/Putin-Orders-Russian-Move-To-GNULinux

[SSJBurger]

Win10 with the TPM disabled so I don't get "upgraded" to 11
I dabble in gaming so that was the most thorny issue, especially since I have a NVIDIA GPU with a G-SYNC monitor. I tried many things including passing the GPU to a VM, but that provided unstable performance in terms of frame pacing due to increased DPC latencies inside the VM. I enjoy competitive FPSes, so it's a deal-breaker. Having to juggle monitors and/or their sources was tiresome as well. Other solutions like native Steam, Proton, Lutris are all mostly fine, but require fiddling from time to time and I'm too lazy for that :P
I also require the full Visual Studio, but that can be fixed in a number of ways, with RDP and/or a VM.
Most other stuff I managed to work around, for example by buying a mouse and keyboard which can be programmed from Linux or store the configuration on device, so it doesn't require a DOS-style resident Cloud App (so not Razer).
All in all, I do want to make the switch in the future, but the current crop of issues prevent it.

I agree with swapping when you are fully ready. On a bright not I have a 2080 Super and Gsync monitor which feels smoother on Linux than Windows also my fps seems to overall stay higher in Linux using linux-tkg pds kernel (recently did a full dive when Apex was supported) but will note i disable my second monitor when gaming... also I use a Razer mouse and keyboard which has a full gui program in Linux openrazer+polychromatic. So that's most of the issues you are having.. mainly I would say not being able to use dual monitors is the huge downside.

[Ironmask]

It wouldn't. One broken garbage is enough. Don't let it spread.

No need to be so harsh, Arch isn't that bad as long as you update frequently and know how to rescue a broken system

[DMJC]

For running Windows software, there is a lot of overlap with WINE (and Valve's Proton for games). I doubt ReactOS will ever overtake those in usefulness.

Sorry mate, it already has. I have multiple software packages that run better in ReactOS than in WINE. Usually due to the ReactOS display model being better than WINE's hacks on X11/Wayland. ReactOS can already run 3D Studio Max, and Caligari trueSpace better than WINE can. TBH though I use Blender now so I don't really have a use case for ReactOS other than some old games that WINE still struggles with.

[cb88]

Here I would tend to disagree. I remember when I was at school (late 90's) that Microsoft was proudly announcing support for up to 16 processors for NT while benchmarks showed that beyond the second CPU, there was no performance improvements any more. At the same time, Linux' support for multi processors was experimental but was aiming at scaling properly up to 64 CPUs (at doing well at 8 cpus). And we got a new supercomputer running Linux with 64 processors around 2002 that got later upgraded to 128 processors (it was an SGI Altix if I remember well, but here the launch dates given on Wikipedia don't match: it says it was launched in 2003 but I left that place in 2002 and I remember the whole installation of this new system... Maybe as we were a military research center we got the system earlier than advertised?)

Your memory might be accurate but the details are acutally a little different and its come up in online discussions on the topic.

Windows 3.1 had vendor specific HAL support for up to 32 CPUs (subject windows edition license also). Around mid 1994 a generic MPS 1.1 HAL was released that supported either 8 or 16 CPUs, but only 2 are enabled in workstation per the license and registry settings you can change this and enable more. So, either people testing the hardware didn't enable the extra CPUs or what is even more likely is they ran into the poor scaling of early x86 SMP which was *bad* not because of software but because the hardware itself scaled badly irrespective of MS. The same exact code executed in a VM today ought to show decent scaling as the hardware today handles this much better and talks to a nice integrated IMC with a very fat link to cache.... SMP >2 CPUs was definitely bandwidth limited at the time. You would have had to run something like prime95 which exercises mainly the cores themselves and not the IMC to show linear speed up on the hardware of the day.

[cb88]

2001: https://en.wikipedia.org/wiki/POWER4

Not a PC. Also, if you include multi chip modules Ross had multiprocessor HyperSparcs in 1996, and sun had the SM52X earlier than that, as well as dual modules for earlier CPUs but its harder to pin a date on those. And I am pretty sure there were various mini and mainframe CPUs that would qualify much earlier than that, certainly.

For the hypersparcs you'd get 2 CPUs in a single module on Colorado 3 and I think a couple other models, and you could install 2 modules for up to quad processing in a standard SS10 or SS20 chassis. Typical clock rates for Ross's dual modules were 125-200mhz depending on the module in question, early dual modules might have been a little slower, Sun's SM52X is 40Mhz x 2... though there wasn't much point to this once you could install an SM81 at 85Mhz as it was probably 2.5-3x faster, as well as being faster clock for clock than any hypersparcs (but not fast enough to beat the 142Mhz+ hypersparcs)