🚀 go-pugleaf

1) AMD make better CPUs than Intel, and are still doing so (IMO).
2) I run a P4 2.8GHz (HT) brand new PC at work and it is about as fast as my
1.4GHz AMD Athlon based PC which is 4 years old now!!
3) Instead of putting #include <math.h> in your application code modules, put
#include <fastmath.h>, and see what the results are then. Please post them up
here again for us to see.
4) BCB6 does not generate HT-friendly code. Try turning off HT in the BIOS for
the P4 PCs, and see what the times are then. Again, let us know what they are.
5) We are waiting to hear whether there will be any updates to BCB6. An
announcement is due on 15th December.

You do not state what flavour of XP you have. I use SP1 and it is definitely
faster than SP2. I investigated as to why this should be, and discovered that
M$ have put loads of security-checking code into the kernel of the OS in SP2,
and this will slow most anything down, but offer improved security (a bit like
a sandbox). I have already written my own tray app called MJ Registry Watcher,
which monitors the system for unsolicited changes to important registry keys
and system files ( http://www.jacobsm.com/index.htm#sft ). So, I uninstalled
SP2 (you have to reinstall SP1 to do this - you'll need the original XP SP1 cd
with your key), applied the only SP1 patch that was ever useful - the RPC
patch KB823980, and run the latest ZoneAlarm with MJ Registry Watcher. I am
untouchable and yet very very fast compared with SP2. What really got me about
SP2 was the increased boot up time - my God, what the heck were they playing
at when M$ thought this one up!?! Perhaps they need to support Intel's
flagging chip sales, by making their OS so slow, everyone has to upgrade their
CPUs!
-- 
Mark Jacobs
DK Computing
http://www.dkcomputing.co.uk

"Matija" <matija@REMOVEtel.fer.hr> wrote in message
news:4199b948$1@newsgroups.borland.com...
| I developed some algorithms in C++ and compiled them using CPP Builder 6.
| It is about extensive calculations, and all algorithms use object oriented
| approach (using objects instead of some static fields etc.).
|
| Performance results are as follows:
|  1. P4 2.8GHz HT, 512MB RAM, WinXP - calc. time - 88 sec
|  2. P4 2.8GHz HT, 512MB RAM, Win2000 - calc. time - 55 sec
|  3. P4 3.2GHz HT, 2GB RAM, WinXP - calc. time - 75 sec
|  4. AMD Athlon 1.8 GHz (2200+), cca 700MB RAM, WinXP - calc. time - 30 sec
|
| Could someone explain why AMD is working so much faster?
| I thought it could be because standard FPU operations are faster on AMD
| (according Sandra benchmarks)?!
|
| As much I understood CPP Builder 6 compiler doesn't support SSE2 instruction
| set (which should increase P4 performances).
|
| Are there any compiler updates that could allow SSE2 instruction set to be
| used?
|
|

I just compiled with fastmath.h.
Since in last day I added some debug variables to my code, it has slown 
down... But, with fastmath.h I didn't gain any speed (at least not on P4 - I 
only tested it on P4 at work).
Today I will try it on Centrino and I will post results to group.

I already tryed with HT turned off, it also didn't work anything faster...


"Mark Jacobs" <http://www.jacobsm.com/mjmsg?Borland%20Newsgroup> wrote in 
message news:419b4ebb$1@newsgroups.borland.com...
> 1) AMD make better CPUs than Intel, and are still doing so (IMO).
> 2) I run a P4 2.8GHz (HT) brand new PC at work and it is about as fast as 
> my
> 1.4GHz AMD Athlon based PC which is 4 years old now!!
> 3) Instead of putting #include <math.h> in your application code modules, 
> put
> #include <fastmath.h>, and see what the results are then. Please post them 
> up
> here again for us to see.
> 4) BCB6 does not generate HT-friendly code. Try turning off HT in the BIOS 
> for
> the P4 PCs, and see what the times are then. Again, let us know what they 
> are.
> 5) We are waiting to hear whether there will be any updates to BCB6. An
> announcement is due on 15th December.
>
> You do not state what flavour of XP you have. I use SP1 and it is 
> definitely
> faster than SP2. I investigated as to why this should be, and discovered 
> that
> M$ have put loads of security-checking code into the kernel of the OS in 
> SP2,
> and this will slow most anything down, but offer improved security (a bit 
> like
> a sandbox). I have already written my own tray app called MJ Registry 
> Watcher,
> which monitors the system for unsolicited changes to important registry 
> keys
> and system files ( http://www.jacobsm.com/index.htm#sft ). So, I 
> uninstalled
> SP2 (you have to reinstall SP1 to do this - you'll need the original XP 
> SP1 cd
> with your key), applied the only SP1 patch that was ever useful - the RPC
> patch KB823980, and run the latest ZoneAlarm with MJ Registry Watcher. I 
> am
> untouchable and yet very very fast compared with SP2. What really got me 
> about
> SP2 was the increased boot up time - my God, what the heck were they 
> playing
> at when M$ thought this one up!?! Perhaps they need to support Intel's
> flagging chip sales, by making their OS so slow, everyone has to upgrade 
> their
> CPUs!
> -- 
> Mark Jacobs
> DK Computing
> http://www.dkcomputing.co.uk
>
> "Matija" <matija@REMOVEtel.fer.hr> wrote in message
> news:4199b948$1@newsgroups.borland.com...
> | I developed some algorithms in C++ and compiled them using CPP Builder 
> 6.
> | It is about extensive calculations, and all algorithms use object 
> oriented
> | approach (using objects instead of some static fields etc.).
> |
> | Performance results are as follows:
> |  1. P4 2.8GHz HT, 512MB RAM, WinXP - calc. time - 88 sec
> |  2. P4 2.8GHz HT, 512MB RAM, Win2000 - calc. time - 55 sec
> |  3. P4 3.2GHz HT, 2GB RAM, WinXP - calc. time - 75 sec
> |  4. AMD Athlon 1.8 GHz (2200+), cca 700MB RAM, WinXP - calc. time - 30 
> sec
> |
> | Could someone explain why AMD is working so much faster?
> | I thought it could be because standard FPU operations are faster on AMD
> | (according Sandra benchmarks)?!
> |
> | As much I understood CPP Builder 6 compiler doesn't support SSE2 
> instruction
> | set (which should increase P4 performances).
> |
> | Are there any compiler updates that could allow SSE2 instruction set to 
> be
> | used?
> |
> |
>
>