> A confusing presentation
Sorry, I'd been up for 20 hours by the time I posted... it shows :-)
Andrew McP
> I thought that a V3 is faster in GPL than GeForce/DDR
No idea, don't have a GeForce. You using full detail mirrors? My numbers
are all full detail. I don't race like that, but it's useful for
benchmarking.
> Q2 crusher, same settings: 78fps.
It's Q3, Barrysworld(.com)'s crusher.
Andrew McP
PC133 RAM, V3@175MHz, ABit BH6 early rev (nv bios)
Silverstone, max detail@1024x768, 19AI, start race at the back of the
field. Let AI drive off without you.
(i) frame rate waiting for flag
(ii) minimum during start
Numbers are all minimum whole fps observed.
Cel 464a (104ish MHz) Slot 1 Piii550e@682 (124MHz)
(i) 18 27
(ii) 14 very briefly 23
The same 27 start, 23 dip applies when I actually race (still at the back
of the field). FR reaches 36 well before the bridge, dips to 23-24 into
T1, then up to 36 again very quickly. All this assumes no crashes of
course, during which FR can still go anywhere. From then on it's solid 36
pretty much all the way round.
Silverstone's not the most CPU intensive track (for instance the start at
Spa doesn't give me 36fps until we're clear of Eau Rouge. It's just the
track I've always used for benchmarking. With the detail slider all the
way to the left the new CPU gives a solid 36 throughout the start. Slight
dips flicker during acceleration and T1, but it's very brief.
With an ABit Be6 II, CPUs from the same batch are doing 800+. I think that
kind of speed may well see something very close to a solid full detail 36.
Only time will tell, my credit card will need a rest for a month or two,
by which time Celeron2's will be tuppence and running at 4GHz or something
:->
Misc benchmarks:
_____________________________________
3DMark2000 1024x768x16
1610 2204
_____________________________________
Speedy 1280x1024x16 (Hercules windows
-thrashing benchmark)
940 1699
(128k->256k cache prob explains this leap)
_____________________________________
CPU Mark
38.5 58.3
_____________________________________
Q3: 1024x768x16, max detail, high geometry
demo001 29.6 36.1
demo002 19.6 38.3
crusher 13.3 30.9
(confirms that the V3's the limit in simple
scenes, CPU's the limit the busier it gets)
_____________________________________
Flanker 2.02: 1024x768, maxxed out detail inc mirrors, 50m over Sevastopol
Minimum: 11 fps, average 12-15
_____________________________________
This post was sponsored by those nice people at VISA.
Andrew McP
***************************************
NOTE: no notionally inferior sims/games
were harmed in the making of this post.
***************************************
A confusing presentation, but thanks for letting us know how much the extra
CPU power affects GPL frame rate with all other factors equal. We can all
look forward to those 1 GHz chips!! I wonder how much difference a Voodoo 4
or 5 will make? I think much less than the type of CPU upgrade you
describe.
Marc.
"Andrew MacPherson" <[email hidden]> wrote in message
news:[email hidden]...
Quoted text (79 lines)
> > PC133 RAM, V3@175MHz, ABit BH6 early rev (nv bios) > > Silverstone, max detail@1024x768, 19AI, start race at the back of the > field. Let AI drive off without you. > > (i) frame rate waiting for flag > (ii) minimum during start > > Numbers are all minimum whole fps observed. > > Cel 464a (104ish MHz) Slot 1 Piii550e@682 (124MHz) > > (i) 18 27 > (ii) 14 very briefly 23 > > The same 27 start, 23 dip applies when I actually race (still at the back > of the field). FR reaches 36 well before the bridge, dips to 23-24 into > T1, then up to 36 again very quickly. All this assumes no crashes of > course, during which FR can still go anywhere. From then on it's solid 36 > pretty much all the way round. > > Silverstone's not the most CPU intensive track (for instance the start at > Spa doesn't give me 36fps until we're clear of Eau Rouge. It's just the > track I've always used for benchmarking. With the detail slider all the > way to the left the new CPU gives a solid 36 throughout the start. Slight > dips flicker during acceleration and T1, but it's very brief. > > With an ABit Be6 II, CPUs from the same batch are doing 800+. I think that > kind of speed may well see something very close to a solid full detail 36. > Only time will tell, my credit card will need a rest for a month or two, > by which time Celeron2's will be tuppence and running at 4GHz or something > :-> > > Misc benchmarks: > _____________________________________ > > 3DMark2000 1024x768x16 > > 1610 2204 > _____________________________________ > > Speedy 1280x1024x16 (Hercules windows > -thrashing benchmark) > > 940 1699 > > (128k->256k cache prob explains this leap) > _____________________________________ > > CPU Mark > > 38.5 58.3 > _____________________________________ > > Q3: 1024x768x16, max detail, high geometry > > demo001 29.6 36.1 > demo002 19.6 38.3 > crusher 13.3 30.9 > > (confirms that the V3's the limit in simple > scenes, CPU's the limit the busier it gets) > _____________________________________ > > Flanker 2.02: 1024x768, maxxed out detail inc mirrors, 50m over Sevastopol > > Minimum: 11 fps, average 12-15 > _____________________________________ > > This post was sponsored by those nice people at VISA. > > Andrew McP > > *************************************** > NOTE: no notionally inferior sims/games > were harmed in the making of this post. > *************************************** >Quoted text (21 lines)
> PC133 RAM, V3@175MHz, ABit BH6 early rev (nv bios) > > Silverstone, max detail@1024x768, 19AI, start race at the back of the > field. Let AI drive off without you. > > (i) frame rate waiting for flag > (ii) minimum during start > > Numbers are all minimum whole fps observed. > > Cel 464a (104ish MHz) Slot 1 Piii550e@682 (124MHz) > > (i) 18 27 > (ii) 14 very briefly 23 > _____________________________________ > > Q3: 1024x768x16, max detail, high geometry > > demo001 29.6 36.1 > demo002 19.6 38.3 > crusher 13.3 30.9you might want to check this out, a chart of tests
http://www.dirkwagner.de/benchmark/en_benchmark/body_en_benchmark.html
fwiw, my 300a@450 8meg thriller machine beats all but the top system there
JM
Andrew MacPherson wrote:
Quoted text (77 lines)
> PC133 RAM, V3@175MHz, ABit BH6 early rev (nv bios) > > Silverstone, max detail@1024x768, 19AI, start race at the back of the > field. Let AI drive off without you. > > (i) frame rate waiting for flag > (ii) minimum during start > > Numbers are all minimum whole fps observed. > > Cel 464a (104ish MHz) Slot 1 Piii550e@682 (124MHz) > > (i) 18 27 > (ii) 14 very briefly 23 > > The same 27 start, 23 dip applies when I actually race (still at the back > of the field). FR reaches 36 well before the bridge, dips to 23-24 into > T1, then up to 36 again very quickly. All this assumes no crashes of > course, during which FR can still go anywhere. From then on it's solid 36 > pretty much all the way round. > > Silverstone's not the most CPU intensive track (for instance the start at > Spa doesn't give me 36fps until we're clear of Eau Rouge. It's just the > track I've always used for benchmarking. With the detail slider all the > way to the left the new CPU gives a solid 36 throughout the start. Slight > dips flicker during acceleration and T1, but it's very brief. > > With an ABit Be6 II, CPUs from the same batch are doing 800+. I think that > kind of speed may well see something very close to a solid full detail 36. > Only time will tell, my credit card will need a rest for a month or two, > by which time Celeron2's will be tuppence and running at 4GHz or something > :-> > > Misc benchmarks: > _____________________________________ > > 3DMark2000 1024x768x16 > > 1610 2204 > _____________________________________ > > Speedy 1280x1024x16 (Hercules windows > -thrashing benchmark) > > 940 1699 > > (128k->256k cache prob explains this leap) > _____________________________________ > > CPU Mark > > 38.5 58.3 > _____________________________________ > > Q3: 1024x768x16, max detail, high geometry > > demo001 29.6 36.1 > demo002 19.6 38.3 > crusher 13.3 30.9 > > (confirms that the V3's the limit in simple > scenes, CPU's the limit the busier it gets) > _____________________________________ > > Flanker 2.02: 1024x768, maxxed out detail inc mirrors, 50m over Sevastopol > > Minimum: 11 fps, average 12-15 > _____________________________________ > > This post was sponsored by those nice people at VISA. > > Andrew McP > > *************************************** > NOTE: no notionally inferior sims/games > were harmed in the making of this post. > ***************************************Hi Marc,
There may be hope. The new Voodoo boards are supposed to be much less
processor dependant, and very fast. Same with the nVidia boards, I think.
Course, on the downside, they're very expensive. More than just a new cpu in
a lot of cases. Time will tell, but I hope it's true. The next time I
upgrade my cpu, I'll have to change to PC133 memory too. Let's hope the hype
is true.
Slot
Marc Collins <[email hidden]> wrote in message
news:Wi1H4.47869$[email hidden]...
> A confusing presentation, but thanks for letting us know how much theextra
> CPU power affects GPL frame rate with all other factors equal. We can all
> look forward to those 1 GHz chips!! I wonder how much difference a Voodoo4
> or 5 will make? I think much less than the type of CPU upgrade you
> describe.
>
> Marc.
In article <[email hidden]>,
[email hidden] (Andrew MacPherson) writes:
Yeah, I have to drop a lot of detail to get good frame rates. It's tough to
figure out what to drop and still keep enough graphics for braking points, etc.
Getting the K62-350 was a financial decision.
I had someone tell me that a K6-3 400 would be faster than a Celeron 500. That
doesn't sound right. Which would you suggest?
Eldred
--
Tiger Stadium R.I.P. 1912-1999
Own Grand Prix Legends? Goto http://gpl.gamestats.com/vroc
Never argue with an idiot. He brings you down to his level, then beats you
with experience...
Remove SPAM-OFF to reply.
Quoted text (15 lines)
>> these numbers STILL seem very slow to me. > >I'm afraid they look about right to me :-( Sorry I can't give you a tweak >to get a few more FPS. Best bet is go back and thump whoever suggested >you get the K6 :-) > >You can see from your numbers that lowering the resolution made v.little >difference, which means the CPU's the limiting factor. > >Can you race at 36fps with much less detail? If you can't then that's one >of the reasons you're still a bit slow on the track. I'll never forget >moving from a P225mmx to a Cel450a... it was like driving a different sim. >Solid frame rates are Your Best Friend. Moving up from 450 to 682 is much >more of a luxury, because it only really allows me to have more detail >(all offline).
> the K62-350 was a financial decision.
I thought so, and can sympathise. I worked out a long time ago I can
afford to run a PC or a car. The PC's still winning, there's less chance
of dying at the keyboard for a start... though some T1 crashes can do
nasty things to my blood pressure<g>.
> I had someone tell me that a K6-3 400 would be
> faster than a Celeron 500.
The only AMD worth having IMO is an Athlon, and if you're springing for a
new motherboard etc then I -- personally -- prefer to tread the mainstream
path. I've always overclocked Intel chips... 486sx25->sx40, P90->P120,
P166->P225mmx, Cel300a->450a, Now 550e->682 (it's "guaranteed" to 733 and
chips from the same batch are doing 800+). Athlons seem to overclock well,
but previous chips have always run as close to their limit as AMD can
risk. Add to that the fact that AMD's only just discovered how to design
their FPU to match (and under some tests, beat) Intel's and I'd steer
clear of older AMDs. Of course if you can slot a faster CPU into the same
motherboard, that might be worth doing, as the only thing that *really*
matters is MHz. But given that a Cel500 has more MHz, and the better FPU,
I'd say you got some lousy advice :-)
Andrew McP
> Don't know if you like N3
Love N3... though only because buying it gives me more GPL tracks to drive
;-)
> I get 2865 in 3DMark2000 - what drivers are you using?
Probably down to my Nov99 (I think) drivers. I tend not to change them
unless I need to. Hopefully the relative gap between the two processors is
valid though.
Andrew McP
Pretty sweet, eh? I've got a nearly identical system as yours, even
the same CPU upgrade (did stop at a PIII 500 in the middle, though).
I've got a V3-2000@172Mhz. Don't know if you like N3, but I get a
consistent 28fps+ @1024x768, all details, 20vis front, 6 behind, with
antialiasing enabled!
I get 2865 in 3DMark2000 - what drivers are you using? Maybe you're
missing out on a system tweak?
Kendt
In article <[email hidden]>,
[email hidden] wrote:
back
> of the field). FR reaches 36 well before the bridge, dips to 23-24into
> T1, then up to 36 again very quickly. All this assumes no crashes of
> course, during which FR can still go anywhere. From then on it'ssolid 36
> pretty much all the way round.
>
> Silverstone's not the most CPU intensive track (for instance thestart at
> Spa doesn't give me 36fps until we're clear of Eau Rouge. It's justthe
> track I've always used for benchmarking. With the detail slider allthe
> way to the left the new CPU gives a solid 36 throughout the start.Slight
> dips flicker during acceleration and T1, but it's very brief.
>
> With an ABit Be6 II, CPUs from the same batch are doing 800+. I thinkthat
> kind of speed may well see something very close to a solid fulldetail 36.
> Only time will tell, my credit card will need a rest for a month ortwo,
> by which time Celeron2's will be tuppence and running at 4GHz orsomething
Sevastopol
Sent via Deja.com http://www.deja.com/
Before you buy.
Quoted text (17 lines)
> > PC133 RAM, V3@175MHz, ABit BH6 early rev (nv bios) > > Silverstone, max detail@1024x768, 19AI, start race at the back of the > field. Let AI drive off without you. > > (i) frame rate waiting for flag > (ii) minimum during start > > Numbers are all minimum whole fps observed. > > Cel 464a (104ish MHz) Slot 1 Piii550e@682 (124MHz) > > (i) 18 27 > (ii) 14 very briefly 23 > > The same 27 start, 23 dip applies when I actually race (still at theQuoted text (34 lines)
> :-> > > Misc benchmarks: > _____________________________________ > > 3DMark2000 1024x768x16 > > 1610 2204 > _____________________________________ > > Speedy 1280x1024x16 (Hercules windows > -thrashing benchmark) > > 940 1699 > > (128k->256k cache prob explains this leap) > _____________________________________ > > CPU Mark > > 38.5 58.3 > _____________________________________ > > Q3: 1024x768x16, max detail, high geometry > > demo001 29.6 36.1 > demo002 19.6 38.3 > crusher 13.3 30.9 > > (confirms that the V3's the limit in simple > scenes, CPU's the limit the busier it gets) > _____________________________________ > > Flanker 2.02: 1024x768, maxxed out detail inc mirrors, 50m overQuoted text (14 lines)
> > Minimum: 11 fps, average 12-15 > _____________________________________ > > This post was sponsored by those nice people at VISA. > > Andrew McP > > *************************************** > NOTE: no notionally inferior sims/games > were harmed in the making of this post. > *************************************** > >
> these numbers STILL seem very slow to me.
I'm afraid they look about right to me :-( Sorry I can't give you a tweak
to get a few more FPS. Best bet is go back and thump whoever suggested
you get the K6 :-)
You can see from your numbers that lowering the resolution made v.little
difference, which means the CPU's the limiting factor.
Can you race at 36fps with much less detail? If you can't then that's one
of the reasons you're still a bit slow on the track. I'll never forget
moving from a P225mmx to a Cel450a... it was like driving a different sim.
Solid frame rates are Your Best Friend. Moving up from 450 to 682 is much
more of a luxury, because it only really allows me to have more detail
(all offline).
Andrew McP... feeling a bit guilty about his big numbers :-)
In article <[email hidden]>,
[email hidden] (Andrew MacPherson) writes:
Ok, using this same criteria...(!):
With Dyn. Light/hilights without
(i) 8 9
(ii) 7 7
In 640x480
(i) 9 10
(ii) 7 8
My system:
K62-350, 128MB, V3 3000(not o/c). Granted, my system isn't even CLOSE in power
as Andrew's, but these numbers STILL seem very slow to me. I'm sure that there
are others with similar systems who get much better rates. How far off am I,
and what suggestions are there...?
Eldred
--
Tiger Stadium R.I.P. 1912-1999
Own Grand Prix Legends? Goto http://gpl.gamestats.com/vroc
Never argue with an idiot. He brings you down to his level, then beats you
with experience...
Remove SPAM-OFF to reply.
Quoted text (14 lines)
>PC133 RAM, V3@175MHz, ABit BH6 early rev (nv bios) > >Silverstone, max detail@1024x768, 19AI, start race at the back of the >field. Let AI drive off without you. > >(i) frame rate waiting for flag >(ii) minimum during start > >Numbers are all minimum whole fps observed. > > Cel 464a (104ish MHz) Slot 1 Piii550e@682 (124MHz) > >(i) 18 27 >(ii) 14 very briefly 23Quoted text (15 lines)
>> I had someone tell me that a K6-3 400 would be >> faster than a Celeron 500. > >The only AMD worth having IMO is an Athlon, and if you're springing for a >new motherboard etc then I -- personally -- prefer to tread the mainstream >path. I've always overclocked Intel chips... 486sx25->sx40, P90->P120, >P166->P225mmx, Cel300a->450a, Now 550e->682 (it's "guaranteed" to 733 and >chips from the same batch are doing 800+). Athlons seem to overclock well, >but previous chips have always run as close to their limit as AMD can >risk. Add to that the fact that AMD's only just discovered how to design >their FPU to match (and under some tests, beat) Intel's and I'd steer >clear of older AMDs. Of course if you can slot a faster CPU into the same >motherboard, that might be worth doing, as the only thing that *really* >matters is MHz. But given that a Cel500 has more MHz, and the better FPU, >I'd say you got some lousy advice :-)> From: [email hidden]
> I get 2865 in 3DMark2000 - what drivers are you using?
> Maybe you're missing out on a system tweak?
Mmm, I was getting 2230 ish with this Piii682e/V3 2k@175 a few weeks ago
when I posted the original info. I've just managed to successfully upgrade
from Win95 osr2 to Win98 and, having just run 3DMark2000 again, I now get
3230. I'd be suspicious that I made a mistake before and it was 3000 not
2000, but it was definitely 2 and I ran lots of tests at various speeds
that all fitted in with my expectations. Watching the tests now some run
*much* faster, particularly the heavily textured coloured "pulsing
ring" thing and the cpu speed test on the heli/medieval town scenes... in
the latter the little blokes were running around like Keystone Cops...
never seen it run like that before!
So I'm puzzled. The PC does feel a bit slicker with W98... but most
benchmarks/sims are pretty much identical.
Andrew McP... looking a bit puzzled
PS It's not down to the Piii or D3D software optimisation choice in 3DMark
either... running in Piii optimised mode gives 3230, in software mode it's
more like 3020.


































