GP2 - Flawed Logic

I felt I would add my two cents worth to all of the discussion on the delays behind GP2. There have been many people who are crapping on about how MP/SH have been waiting until faster chips come out, what a load of crap! It has been well documented that the new Pentium Pro machines DO NOT run 16 bit code any faster than the current Pentiums. Unless you are running the game on a true 32 bit operating system (NT or OS/2) than there is no gain. Windows 95 is not a true 32 bit operating system, it has large amounts of 16 bit code below it. It is already documented that Windows 95 runs slower on some new Pentium Pros. The game is written using 16 bit code, although it has a flat memory model thanks to dos4gw. The code is 16 bit, the OS is 16 bit, so a Pentium 133 etc is the best you will get. If you want any faster performance than buy a better video card and get more system memory. Like many others, I am patiently awaiting the release date! Grant. //// (0 0) ============================oOOo==(_)==oOOo================================== PHONE +61 7 3247 4968 / * FAX +61 7 3247 5033 \_,-._/ Brisbane,Queensland,Australia. (DOS one day, NT the next)! ============================================================================= ooO Ooo

Well, a P-166, P-180 and P-200 will all be significantly faster than a P-133, and a P-150 may well be faster too, although only slightly if it is. FWIW, Marc (US half of Sim Racing News) had a play with GP2 on a Dell 166 at the E3 show. The P-166 will shift GP2 at....10fps with ALL detail on. Switch off asphalt and clouds and it pretty much motors. Drop a bit of mirror detail and you're into 120/133 territory, drop..... You get the picture - we'll all be able to run it at some configuration, with even nicer stuff to look forward to as hardware develops. Cheers! John _________________________________ __ _____| |_____ __ _________| |__| :| John Wallace | |__| |_________ \ :| |::| :| Team WW Racing TSW | |::| | / > :| |::| :|_________________________________| |::| | < / :|__|::|____:/ Sim Racing News \.____|::|__| \ /_______:/ \::/ http://www.dcn.ed.ac.uk/pulse/index.htm \::/ \._______\

Quoted text (24 lines)

>I felt I would add my two cents worth to all of the discussion on the >delays behind GP2. There have been many people who are crapping on about >how MP/SH have been waiting until faster chips come out, what a load of crap! >It has been well documented that the new Pentium Pro machines DO NOT run >16 bit code any faster than the current Pentiums. Unless you are running >the game on a true 32 bit operating system (NT or OS/2) than there is no >gain. Windows 95 is not a true 32 bit operating system, it has large >amounts of 16 bit code below it. It is already documented that Windows >95 runs slower on some new Pentium Pros. >The game is written using 16 bit code, although it has a flat memory >model thanks to dos4gw. The code is 16 bit, the OS is 16 bit, so a >Pentium 133 etc is the best you will get. If you want any faster >performance than buy a better video card and get more system memory. >Like many others, I am patiently awaiting the release date! >Grant. > //// > (0 0) >============================oOOo==(_)==oOOo================================== >PHONE +61 7 3247 4968 / * >FAX +61 7 3247 5033 \_,-._/ >Brisbane,Queensland,Australia. (DOS one day, NT the next)! >============================================================================= > ooO Ooo

How about a P5 166 or perhaps a P5 200 or maybe even P200 with MMX. The P5 is still being developed and M/P might well be waiting for the _average_ machine to be a P133. Mike

>How about a P5 166 or perhaps a P5 200 or maybe even P200 with MMX. The game would have to be written to utilise those MM instructions of course. Or it would have to test the CPU, and implement it's own system if they are not available. I don't know how useful those MM instructions will be to game developers. Also, looking towards MMX is a bit like waiting on a 3DCard standard. It's difficult to write a game to use a particular card and ignore others, or to support all cards on the market. Is there a 3D standard emerging yet? This isn't something I've taken too much interest in yet. With MMX, the CPU's won't be available (ready?) until the end of the year I thought. Entry level is a P100 these days.... Stuart -- Stuart Booth Somewhere in Godalming, England, UK

Crap, DOS/4GW executables are 32 bit protected mode executables running 32 bit instructions - using 32 bit registers, using 32 bit protected mode interrupt handlers. Memory access is 32 bit (when required), and bus access can be 32 bit when the hardware is 32 bits (PCI/VESA rather than ISA video cards). When running on DOS, the OS is 16 bit, but the extent of OS interaction is file system access, and VESA SVGA capabilities detection at initialisation time. When running on OS/2 or NT (or any real 32 bit operating system with decent DPMI support), operating system calls such as INT 21 file access can be handled without ever leaving protected mode. So kindly learn the facts before you fill this newsgroup with your crap. F1GP2 will run faster on a Sexium (Pentium Pro), but I firmly believe that God (Geoff Crammond) will make a game that far exceeds anything you have seen before on equivalent hardware. -- | // There is always one more bug! | | // Windows 95 : Busy or Unstable? Don't ask me, I run OS/2 |

=> When running on OS/2 or NT (or any real 32 bit operating system with => decent DPMI support), operating system calls such as INT 21 file access => can be handled without ever leaving protected mode. So a P6 will run game faster than P5 if the OS is NT, and P6 will run game slower than P5 if the OS is Win95. Is this right? => So kindly learn the facts before you fill this newsgroup with your crap. => => F1GP2 will run faster on a Sexium (Pentium Pro), but I firmly believe that => God (Geoff Crammond) will make a game that far exceeds anything you have => seen before on equivalent hardware. => => -- => | // There is always one more bug! | => | // Windows 95 : Busy or Unstable? Don't ask me, I run OS/2 |

>=> When running on OS/2 or NT (or any real 32 bit operating system with >=> decent DPMI support), operating system calls such as INT 21 file access >=> can be handled without ever leaving protected mode. >So a P6 will run game faster than P5 if the OS is NT, and P6 will run game >slower than P5 if the OS is Win95. Is this right? No - because it wouldn't be making operating system calls during the game, unless there is a record lap to disk option (like nascar), but then that should still be pretty insignificant. Anyway Win95 may be able to do disk I/O from a DPMI client without switching to real/v86 mode. Besides it also depends on the hardware virtualisation provided by the operating system. For example, enabling/disabling interrupts are protected instructions in V86 mode, and thus cause a trap into protected mode where the DOS emulator will simulate the operation. The same goes for hardware I/O like to the sound card. Some things can run faster when using virtualised hardware - for instance I/O to an ISA bus video card will speed up (higher bandwidth) during background/windowed execution, because the virtual (simulated) video device will in fact be high speed main memory. The other issue to effect P6 execution speed is code optimisation to keep the instruction pipeline full (effectively arranging the instructions in the appropriate order so that they can be executed in parallel). If GP2 was already optimised to take advantage of this (not disadvantage) on a P5 then I suspect it would run just as effectively on a P6 (with little or no pipeline stalls). But I expect that since the program is written in assembly by Geoff, rather than a "highly" optimised compiler it is unlikely to use this optimisation technique - and might only get a 0.1% average speedup (your guess is probably better than mine?). -- | // There is always one more bug! | | // Windows 95 : Busy or Unstable? Don't ask me, I run OS/2 |

>How about a P5 166 or perhaps a P5 200 or maybe even P200 with MMX. >The P5 is still being developed and M/P might well be waiting for the >_average_ machine to be a P133. >Mike Perhaps the Cyrix M2 family will grab people's attention. - Pentium Pro performance on 32-bit code, far better on 16-bit code. - 32-bit AND 16-bit code optimization - MMX technology - Use in any 6x86 compatible motherboard (many Pentium motherboards are 6x86 capable). (http://www.cyrix.com/process/prodinfo/6x86/6x-mblst.htm)

Hmmmm, considering it's not here yet, "far better on 16-bit code" is a bit speculative wouldn't you say? Do you really think Intel would license their MMX code to a competitor if they were the slightest bit concerned by it? Another one for silicon heaven methinks. I hope not, about time Intel had some competition at the high-end of the market (where the money is made) but I fear not. Cheers! John _________________________________ __ _____| |_____ __ _________| |__| :| John Wallace | |__| |_________ \ :| |::| :| Team WW Racing TSW | |::| | / > :| |::| :|_________________________________| |::| | < / :|__|::|____:/ Sim Racing News \.____|::|__| \ /_______:/ \::/ http://www.dcn.ed.ac.uk/pulse/index.htm \::/ \._______\

Let's get some basic facts right: PPro's run 32-bit code faster than Pentiums (at the same and clock speeds). Pentiums run 16-bit code faster than PPro's (at the same clock speed). It may be the case that a PPro X MHZ runs 16-bit code faster than a P Y MHZ due to the fact that X is significantly greater than Y (this is not the case for all the current speeds) - anyway, ignore this little bit for now. No it will not. I doubt very much F1GP2 will even run under a NT DOS session, it will access the hardware and other such low level functions that NT will simply not allow. Anyway, just 'cos it's running under NT doesn't mean it's 32-bit, it's still a 16-bit app for a 16-bit OS (DOS). F1GP2 will NOT run faster on a PPro than on a Pentium (given my top statement). F1GP2 is a DOS 16-bit application (even though it will probably use some kind of 32-bit flat memory model like most DOS games do now). Consequently, the fastest way of running F1GP2 will be in plain old DOS, nothing else. The current fastest processor to run it on is a Pentium 166 MHZ - soon to be improved upon with the Pentium 200 MHZ. It should be noted (and most game magazine I've seen recently have run tests to prove it), a P166 (fastest Pentium) is vastly faster than a PPro200 (fastest PPro) for DOS based games. We are talking now Doom2 frame rates of 50+ per sec to ~ 10 per sec. Get a brain before making statements like this, that way you may actually be able to learn the facts. You're not quite that dumb though, as I too am certain F1GP2 will be the most amazing PC game there is. Jamie

>No it will not. I doubt very much F1GP2 will even run under a NT DOS session, it >will access the hardware and other such low level functions that NT will simply >not allow. Anyway, just 'cos it's running under NT doesn't mean it's 32-bit, >it's still a 16-bit app for a 16-bit OS (DOS). We are talking in theory, but you would have to assume that eventually NT will provide decent DOS emulation so that even games will run. But if theory is beyond you then talk about GP2 running on OS/2. Hold on, why did you start calling the application a 16 bit app. If it is using 32 bit registers, 32 bit instructions, and accessing memory in 32 bit chunks (where possible) it is a 32 bit application regardless of the memory model (but that happens to be 32 bit flat memory model). But 32 bit programs can still use 16 bit instructions, it is up to the compiler optimiser to choose the instructions to use. I am sure that NT still has frequent 16 bit instructions in the code - not all data being manipulated is 32 bits. If you disagree with my definition of a 32 bit program then please tell the world your definition. Any technical report can be flawed or biased, and all should be read with this in mind. A technical report from a game magazine shouldn't be trusted at all. 1. What is Doom2 optimised for, 386, 486 or Pentium? I would doubt that it would be optimised for Pentium pipelining. Maybe Doom2 does contain many 16 bit instructions (it is still a 32 bit program though). 2. Was the hardware setup the equivalent. I am sure there are plenty of bogus PPro motherboard chipsets, simply because it is newer. Was the video card the same? Assuming the results were accurate what did they do to explain them - did they try running NT on the two machines and test for any weird performance problems. It is just too far fetched to expect us to believe that a PPro200 is less than one fifth of the speed of a P166 even when executing fully 16 bit code. 3. Do you really think the PPro is that pathetic? No I haven't used one, but they cost a fortune and some people still buy them - because they need the higher performance. 4. The PPro is probably (I don't know the facts) much faster than a Pentium when it comes to floating point math (as opposed to just a bit faster for integer math). Doom2 would not have used floating point (it runs on a 386), GP2 may not use floating point either - but it is another thing to keep in mind when doing PPro/Pentium comparisons. Mark -- | // There is always one more bug! | | // Windows 95 : Busy or Unstable? Don't ask me, I run OS/2 |

Quoted text (18 lines)

>Subject: Re: GP2 - Flawed Logic >Date: 25 May 1996 19:43:05 GMT >>How about a P5 166 or perhaps a P5 200 or maybe even P200 with MMX. >>The P5 is still being developed and M/P might well be waiting for the >>_average_ machine to be a P133. >>Mike > Perhaps the Cyrix M2 family will grab people's attention. >- Pentium Pro performance on 32-bit code, far better on 16-bit code. >- 32-bit AND 16-bit code optimization >- MMX technology >- Use in any 6x86 compatible motherboard (many Pentium motherboards > are 6x86 capable). (http://www.cyrix.com/process/prodinfo/6x86/6x-mblst.htm)

Actually, based on problems I've had with some games, most recently Sierra products, I'd avoid the Cyrix chip. It's incompatible with some software... ________ Eldred Pickett I am NOT paranoid. And why are you always watching me?!?

Quoted text (39 lines)

> Let's get some basic facts right: PPro's run 32-bit code faster than Pentiums > (at the same and clock speeds). Pentiums run 16-bit code faster than PPro's (at > the same clock speed). It may be the case that a PPro X MHZ runs 16-bit code > faster than a P Y MHZ due to the fact that X is significantly greater than Y > (this is not the case for all the current speeds) - anyway, ignore this little > bit for now. > >=> When running on OS/2 or NT (or any real 32 bit operating system with > >=> decent DPMI support), operating system calls such as INT 21 file access > >=> can be handled without ever leaving protected mode. > >So a P6 will run game faster than P5 if the OS is NT, and P6 will run game > >slower than P5 if the OS is Win95. Is this right? > No it will not. I doubt very much F1GP2 will even run under a NT DOS session, it > will access the hardware and other such low level functions that NT will simply > not allow. Anyway, just 'cos it's running under NT doesn't mean it's 32-bit, > it's still a 16-bit app for a 16-bit OS (DOS). > >=> F1GP2 will run faster on a Sexium (Pentium Pro), but I firmly believe that > >=> God (Geoff Crammond) will make a game that far exceeds anything you have > >=> seen before on equivalent hardware. > F1GP2 will NOT run faster on a PPro than on a Pentium (given my top statement). > F1GP2 is a DOS 16-bit application (even though it will probably use some kind of > 32-bit flat memory model like most DOS games do now). Consequently, the fastest > way of running F1GP2 will be in plain old DOS, nothing else. The current fastest > processor to run it on is a Pentium 166 MHZ - soon to be improved upon with the > Pentium 200 MHZ. > It should be noted (and most game magazine I've seen recently have run tests to > prove it), a P166 (fastest Pentium) is vastly faster than a PPro200 (fastest > PPro) for DOS based games. We are talking now Doom2 frame rates of 50+ per sec > to ~ 10 per sec. > >=> So kindly learn the facts before you fill this newsgroup with your crap. > Get a brain before making statements like this, that way you may actually be > able to learn the facts. You're not quite that dumb though, as I too am certain > F1GP2 will be the most amazing PC game there is. > JamieYou really should get your facts straight, a Pentium Pro 150 is about 5%

faster in 16 bit code than a P166. This is often described as terrible because it is an expensive next generation chip and it only provides a 5% 16bit speed boost.

MS> >No it will not. I doubt very much F1GP2 will even run under a NT MS> DOS session, it will access the hardware and other such low level MS> functions that NT will simply not allow. Anyway, just 'cos it's MS> running under NT doesn't mean it's 32-bit, it's still a 16-bit app MS> for a 16-bit OS (DOS). MS> We are talking in theory, but you would have to assume that MS> eventually NT will provide decent DOS emulation so that even games MS> will run. But if theory is beyond you then talk about GP2 running MS> on OS/2. Well, F1GP runs VERY nicely under OS/2 - faster than it does under DOS on my machine. The Papyrus sims are much the same - no problems and with memory problems defeated (all my sessions run with 640kb base memory, no hassle with TSR's drivers, etc) all seems well. Typically OS/2 in fact! If software is making "illegal" calls to the hardware perhaps the poor software should be faulted, not the OS..... Cheers! --- * RM 1.3 U0414 * I don't do Windows, but OS/2 does.

A typical OS/2 user I see - there must be an incredible amount of "poor" software around, since it's never the fault of OS/2. Funny, I never have any trouble at all running that "poor" software under DOS, could it be that OS/2 just can't emulate DOS properly? Nah, couldn't be that could it..... Cheers! John