Intel's 10nm only coming to servers in 2020 with Ice Lake

[S w a t s o n]

10 minutes ago, M.Yurizaki said:

Of course backporting DX12 back to Windows 3.1 wouldn't work: Microsoft changed the graphics driver model in Windows Vista. This is the same reason why they couldn't backport DX10 to XP (any "DX10 for XP" packages are likely hacks and wrappers to translate DX10 calls to DX9 calls or something). At the moment, I don't see a reason Microsoft can't backport DX12 to Windows 8 or 7 other than they want to drive Windows 10 adoption.

 

If you give it a driver that is compatible with the OS's driver model, it'll know what to do. Windows 7 doesn't know anything about USB 3.0, but you can give a USB 3.0 controller driver and suddenly it's happy with USB 3.0.

 

Similarly Windows XP wouldn't know what to do with a GeForce GTX 480. But give it a driver and suddenly it knows what to do with it.

 

The OS doesn't pass instructions to the CPU. The OS tells the application it can run on the CPU and the application goes from there. The only time the application would ever ping the OS is if a system call needs to be made (i.e., using the graphics card driver). By that point, the application already made its draw calls so it can tell the driver where to look for them.

 

If you're passing deep copies of data around between threads, processes, or whatever without a good reason, you're doing it wrong.

So besides the fact that you literally couldn't get the OS to run on a modern CPU without a rewrite (no getting around that),

 

Simply handing the OS a driver for the GPU isnt enough. Besides the fact that GPU drivers may as well be separate from a typical device driver in their complexity and implementation, the OS would not understand the instruction sets used by these games because of the CPU issue.

 

It sounds nice to say you can do this in theory but I think if you paid Nvidia $1 billion to do it (without rewriting the OS) they would just laugh at you.

 

Also why are you agreeing with me that backporting DX12 to Windows 3.1 wouldn't work but assuming you could somehow do it for OS/2?

[Mira Yurizaki]

18 minutes ago, Swatson said:

So besides the fact that you literally couldn't get the OS to run on a modern CPU without a rewrite (no getting around that),

You could most certainly get OS/2 to run on a modern CPU because it's an x86 based OS, as long as it's not relying on undocumented or deprecated behavior like the A20 Line issue with DOS applications built for the original PC running on something newer. Now if you want it to take advantage of features on modern processors, then sure, you need a rewrite.

 

I mean, LGR got MS-DOS to run on a Ryzen machine. And not just "oh cool, it boots", but run actual software as well.

Quote

Simply handing the OS a driver for the GPU isnt enough. Besides the fact that GPU drivers may as well be separate from a typical device driver in their complexity and implementation, the OS would not understand the instruction sets used by these games because of the CPU issue.

The GPU driver's job is to accept some form of precompiled instructions from some API (which theoretically as long as you compile it to some form that the driver accepts, the actual API doesn't matter, which is how people got Doom 3 running on Voodoo2) and compile it to whatever instructions the GPU wants. This isn't any different from any other driver. Actually, this is literally what drivers do.

Quote

It sounds nice to say you can do this in theory but I think if you paid Nvidia $1 billion to do it (without rewriting the OS) they would just laugh at you.

I think NVIDIA would consider it if you actually had the money. After all, the IRS paid Microsoft for Windows XP support after Microsoft declared it deader than dead.

Quote

Also why are you agreeing with me that backporting DX12 to Windows 3.1 wouldn't work but assuming you could somehow do it for OS/2?

Because we weren't using backporting DX12 to OS/2 as an example? Or I was on a single track mind when I wrote that.

[S w a t s o n]

4 minutes ago, M.Yurizaki said:

You could most certainly get OS/2 to run on a modern CPU because it's an x86 based OS, as long as it's not relying on undocumented or deprecated behavior like the A20 Line issue with DOS applications built for the original PC running on something newer. Now if you want it to take advantage of features on modern processors, then sure, you need a rewrite.

 

I mean, LGR got MS-DOS to run on a Ryzen machine. And not just "oh cool, it boots", but run actual software as well.

The GPU driver's job is to accept some form of precompiled instructions from some API (which theoretically as long as you compile it to some form that the driver accepts, the actual API doesn't matter) and compile it to whatever instructions the GPU wants. This isn't any different from any other driver. Actually, this is literally what drivers do.

I think NVIDIA would consider it if you actually had the money. After all, the IRS paid Microsoft for Windows XP support after Microsoft declared it deader than dead.

Because we weren't using backporting DX12 to OS/2 as an example? Or I was on a single track mind when I wrote that.

Running MS-DOS is not running a 3d game. We're talking about workloads miles apart. And yes to run a 3d game it would need those features requiring a rewrite.

The IRS paying Micro$haft for continued updates is a completely different situation. Additional support is not trying to make new hardware and software run on an old OS.

 

We were talking about OS/2 running modern games (Crysis) that's what started this whole convo

[S w a t s o n]

Remember the shitstorm when Windows 10 was "required" for Kaby-Lake and above due to them not updating Windows 7 to support them. Imagine that but about 1000x worse. Yes it was possible to force Windows 7 to make it work but we're not talking about windows 7 vs windows 10, we're talking about an OS designed when modern, multicore CPUs weren't a thing

[Mira Yurizaki]

2 minutes ago, Swatson said:

Running MS-DOS is not running a 3d game. We're talking about workloads miles apart. And yes to run a 3d game it would need those features requiring a rewrite.

That was a example to counter your point that OS/2 would "literally" not run on a modern CPU. MS-DOS runs just fine on a modern CPU. I would expect OS/2 to run on a modern CPU as well, provided again, it's not relying on undocumented or deprecated behavior.

2 minutes ago, Swatson said:

The IRS paying Micro$haft for continued updates is a completely different situation. Additional support is not trying to make new hardware and software run on an old OS.

Yes it is. Why do you think Intel is still making MS-DOS drivers for GbE NICs? For shits and giggles?

2 minutes ago, Swatson said:

We were talking about OS/2 running modern games (Crysis) that's what started this whole convo

I'm aware of that. You're arguing it's impossible. I'm arguing it is possible. However I'm not arguing it's practical considering how much in-depth knowledge you need to know.

[S w a t s o n]

8 minutes ago, M.Yurizaki said:

That was a example to counter your point that OS/2 would "literally" not run on a modern CPU. MS-DOS runs just fine on a modern CPU. I would expect OS/2 to run on a modern CPU as well, provided again, it's not relying on undocumented or deprecated behavior.

Yes it is. Why do you think Intel is still making MS-DOS drivers for GbE NICs? For shits and giggles?

I'm aware of that. You're arguing it's impossible. I'm arguing it is possible. However I'm not arguing it's practical considering how much in-depth knowledge you need to know.

It's not possible if you can't backport the DirectX API... or in this case OpenGl or whatever the fuck OS/2 even used, probably some proprietary shit aka there is no newer version. Can't port directx to non windows iirc anyways

A fucking gbe nic driver isnt the same as an entire 3d rendering stack, cmon

 

I shoulda quit while I was ahead, RIP intel 10nm thread

[Mira Yurizaki]

36 minutes ago, Swatson said:

It's not possible if you can't backport the DirectX API...

Now that I mentioned it... You don't necessarily need to backport DirectX. You just need to find a way for the application to believe its draw calls are being accepted, intercept them, and turn them into something that the host side can accept. Which again is how people got Doom 3, an OpenGL game, to run on a Voodoo2: they used a Glide wrapper to turn OpenGL calls into Glide ones that the Voodoo2 can understand. And this is how emulators of more recent consoles work. Those consoles use some API, but it's obvious that the API wasn't ported to the host OS.

Quote

A fucking gbe nic driver isnt the same as an entire 3d rendering stack, cmon

A 3D stack that the OS doesn't care about either. Again, all a driver is doing is accepting commands through a standardized interface and turn that into commands the hardware understands. Nothing more.

 

I mean, to go beyond this, we have old computers that don't know what flash memory is accepting CF and SD cards because there's something there to turn the commands the cards use to commands the computers use and vice versa. And I believe something like an SD card reader for the C64, the SD card is still formatted in FAT32, something the C64 has no idea what it is.

 

But yeah, I guess we dragged this on long enough.

[Stefan Payne]

14 hours ago, Swatson said:

I know of those, but they are built upon OS/2 in the same way that Windows has gone through multiple iterations. Not quite as drastic, but I'm talking classic OS/2 to OS/2 Warp, the last official release as of 2001 (ok last off-topic post about things not related to intel 10nm i swur)

 

Edit: ArcaOS wikipedia has this to say about it
 

 

The Graphics Driver was removed after OS Warp 4 and replaced with the VESA Driver you mentioned because that is not something that was needed.

OS/2 became the eCommerce Station and was used up until mid to end 2000s in Banks and other areas.

 

Well, its a shame that it didn't really take off.

But I think its possible that IBM is partly to blame as well with probably less than competitive Pricing for the Software.

 

8 hours ago, M.Yurizaki said:

Linux's driver model as far as I know hasn't changed much since those yesteryear days either, during those dark, dark days.

It did change for some things, especially graphics drivers.
So even with Linux, the usage of an ATi X1800 is pretty limited...

 

5 hours ago, Swatson said:

Backporting DX8 to windows 98 /=/ backporting DX9-DX12 to Windows 3.1 which is basically what you are asking. There are limits. We're talking about an OS that wouldn't know what to do with even an old voodoo card.

Yes, but Direct X9 was the last one that also ran on Windows 9x

And it seems that Crysis can run on Windows 98.

And I've heard that it was possible to start Windows 95 from OS/2, at least I think I heard that someone got it running...

 

5 hours ago, Swatson said:

The graphics API has to run on the OS, the OS passes the instructions to the CPU. The sheer amount of data going through scheduler would break it. Additionally the OS would not be able to run on a modern enough cpu to run the games either, without a rewrite

You are partly wrong here.

With some older OS, it is entirely possible to use them with modern Hardware, because they were less complex and modern CPUs are still compatible to older Code (well, theoretically).

So it should be possible to boot a Windows 9x or maybe even earlier NT with a Ryzen System, if you have one with a PS/2 Connector (well, maybe DOS Emulation works for those)...

 

The Problem is that older OS like Windows 9x only Support 1 CPU, Windows 2000 only supports 2 Cores, for 4 you need Server, w/o SMT you need Advanced Server for a Ryzen and Datacenter for all 32 Cores...

Because Windows 2000 doesn't (or didn't) distinguish between logical and physical processors as they weren't a thing (they started around 4-5 Years Later, in 2004 or 5). 

And AFAIR Windows XP was the first that distinguishes between Physical and Logical processor...

 

5 hours ago, M.Yurizaki said:

You could most certainly get OS/2 to run on a modern CPU because it's an x86 based OS, as long as it's not relying on undocumented or deprecated behavior like the A20 Line issue with DOS applications built for the original PC running on something newer. Now if you want it to take advantage of features on modern processors, then sure, you need a rewrite.

The Software could run on the CPU, the Problem is the Chipset and the I/O Stuff.

While the old Software expects a PS/2 Port and AFAIR OS/2 doesn't support USB, you have kinda a Problem here as there even are Ryzen Boards without a PS/2 Port.

 

But even then, why would you do that?? 

Because of Driver Issues, you won't have much fun anyway. While there are PCIe Graphics cards that have earlier Windows Support.

The Last ATi Card with Win9x Support iis the X-Series so at least it is theoretical possible.

And it should also be possible to boot Windows 9x on an AHCI Controller, I think (not tested it and never will)

 

Windows 2000 and XP however need a Driver for the Disc Controller...

And there is none for Ryzen or the Promontary Chipset so you need a PCIe JMicron PCIe Controller with P-ATA as well and hope that it works... 

5 hours ago, M.Yurizaki said:

I mean, LGR got MS-DOS to run on a Ryzen machine. And not just "oh cool, it boots", but run actual software as well.

Well, to be honest, DOS is a pretty dumb OS and not even that.

It just communicates with the BIOS and lets you do stuff. So most of the actual OS is on the Motherboard anyway.

That is why it runs on modern Hardware...

 

Although most of the DOS compatible Hardware doesn't work with that so you won't have any sound.

And modern Graphics cards also don't support VESA Graphics Modes very well either. I think I tried that a couple of Years ago with a more modern System and it pretty much failed miserabely...

 

5 hours ago, M.Yurizaki said:

Because we weren't using backporting DX12 to OS/2 as an example? Or I was on a single track mind when I wrote that.

OS/2 by the way should have OpenGL Support anyway.
At least up to Warp 4 of course.

Don't know if it was ever used though. Wasn't really a thing at the time...

3DFX Voodoo Graphics only came out in 1996 and everything prior to that was pretty much garbage and therefore the chips before that were even called "decelerator"

 

OS/2 Warp 4 came out a bit before that...

[whitepines]

On 7/29/2018 at 10:12 PM, Dylanc1500 said:

I got Portal 2 to work on "my totally personal not work owned" Power9 workstation with through Linux. It was a very rough experiment with some different drivers.

How did you manage that at a playable frame rate?  Qemu with user mode emulation or something else?

[yian88]

How the hell did we get so fast to a situation flipover, now intel is lagging behind and AMD is leading, what the fucks.

I guess all those RedTeam prayers to the red demon lord payed off, and intel burried itself.

Guys no one realized this yet(or didnt mention it) but intel is going to release dGPU's aswell... on 14nm???!! in 2020+?  both AMD and nvidia are going to have 7nm TSMC GPU's by Q2/Q3 2020 and intel is going to compete in 14nm+++++++++?

Seems like intel set themselves up for bankrupcy, from the greatest fabs and chip maker to nobody, im afraid to consider buying CPU's from them now.

 

Hopefully they manage to get 10nm ready for 2020 it doesnt make sense that they dont, even if they had trouble i doubt they cant make a slightly less dense 10nm but still much better than 14nm just to get some decent competing products, we definately dont want AMD to feel cocky and raise prices cause intel cant catch up.

[cj09beira]

11 minutes ago, yian88 said:

How the hell did we get so fast to a situation flipover, now intel is lagging behind and AMD is leading, what the fucks.

I guess all those RedTeam prayers to the red demon lord payed off, and intel burried itself.

Guys no one realized this yet(or didnt mention it) but intel is going to release dGPU's aswell... on 14nm???!! in 2020+?  both AMD and nvidia are going to have 7nm TSMC GPU's by Q2/Q3 2020 and intel is going to compete in 14nm+++++++++?

Seems like intel set themselves up for bankrupcy, from the greatest fabs and chip maker to nobody, im afraid to consider buying CPU's from them now.

 

Hopefully they manage to get 10nm ready for 2020 it doesnt make sense that they dont, even if they had trouble i doubt they cant make a slightly less dense 10nm but still much better than 14nm just to get some decent competing products, we definately dont want AMD to feel cocky and raise prices cause intel cant catch up.

thats exactly what the rumors say they will do, reduce density a little bit to improve yields, its just sad that we wont get glo fo's 7nm with zen 2, as competition would be even more fierce that way (they had a slightly more advanced node with cobalt for the interconnects and what seemed to be higher clocks, still if amd launches 12 zen 2 cores on 7nm intel wont be able to respond for a while, specially on the server side, amd will be pushing hard to get as many zen 2 cpus sold on the server side as they can, and something tells me it wont be hard to do

[ryao]

On 7/26/2018 at 4:18 AM, Taf the Ghost said:

From the publicly presented information, the total density between the Intel 10nm and GloFo 7nm are roughly similar, though it seems like Intel's cache density will be worse, which is the biggest part of any modern CPU. Thus, the same designs should be slightly smaller on GloFo/TSMC than on the Intel node. The main issue, right now, is if Intel ever releases a 10nm part.

 

Intel is currently roadmapping 4 years behind schedule.

Global Foundaries cancelled their 7nm process and AMD announced that they will be using TSMC.