WinUAE speed changes!

[SHADES]

Hi all,

Just playing around with WinUAE and I have found that it runs 80% faster on my work HP Pentium 4 2.0Gig CELERON computer with onboard Intel crap graphics Windows XP and 512 MB RAM than it does on my ATHLON 2400+ computer with ATI 8500/128 graphics card and also 512MB RAM Windows 2000.

How did I get 80%?? Well I ran AIBB tests on the PC at work as the PC at home seemed less responsive than the one at work. Things like beach ball tests etc. The work one gave 112% bigger than the one at home!!

JIT and all other stuff in enabled, infact I use the same config at home as to the one at work so I don't get it.


What's the go here! my Athlon computer is certainly the faster computer and the graphics hardware would be twice if not more the power of the crappy Intel chipset one.

Is it the fact I am running windows 2000 on the Athlon? I wouldn't have thought so, or is it because some bright spark compiled WinUAE for Pentium 4 specificly?

I'd sure like to run the program faster than at work, but not at the cost of buying a Pentium 4 or having to buy a CRAP Intel graphics card. :pissed:

[SHADES]

@minator

No, I run Windows 2k at home. I woudn't have thought it would be such a difference though, afterall, Windows XP is built on Windows 2000 and I have all the patches installed.
However if Windows XP has THAT much of a perfomance gain on Windows 2000, I'll buy it!!!

[SHADES]

@ Blobrana

I thought P4 CELERON had a lot less CPU Cache than normal P4s??

Remember I did say p4 CELERON.
Surely my ATHLON 2400+ has more cache than the Celeron CPU.

:edit:
Sorry that may have sounded a bit harsh, I should have worded that differently and I do appreciate your effort helping. I'm at a loss to why one is so much faster than the other, well other than the creator compiled WinUAE only for P4 or something like that.

[SHADES]

@ illwrath

I don't know!, to me I would have thought the bigger on die Cache is going to have the greatest perfomance increase over FSB memory access speed in this case as it's where all the common instructions will execute from. the more cache, the mor instructions able to be accessed without the need to re-calculate.

All drivers are up to date for my mainboard including bios.

In all the examples above, no matter what Athlon chip is there, it still has a much bigger on die CPU cache like you said.

[SHADES]

@ Trev

Well that's a real Plus for Intel. Gee I hope you are wrong, I don't like INTEL CPU's

That's nasty coding. Why not just optimize the JIT for standard x86 that is used by all x86 CPUs??

[SHADES]

@ all
Ok, hi people,

here's the reply..

Spec on my home PC
Sandra Specs on CPU
--start--
Generation : G7
Name : Duron M8 & Athlon MP/XP (Thoroughbred) 130nm 1.5-2.5GHz+ 1.5-1.65V
Revision/Stepping : 8 / 1 (0)
Stepping Mask : B0
Core Voltage Rating : 1.650V
Maximum Physical / Virtual Addressing : 34-bit / 32-bit
Native Page Size : 4kB

Chipset 1
Model : VIA Technologies Inc VT8366/A,VT8367 Apollo KT266/A,KT333 CPU to PCI Bridge
Bus(es) : ISA AGP PCI IMB USB FireWire/1394 i2c/SMBus
Front Side Bus Speed : 2x 136MHz (272MHz data rate)
Maximum FSB Speed / Max Memory Speed : 2x 166MHz / 2x 166MHz
Width : 64-bit
IO Queue Depth : 4 request(s)

Logical/Chipset 1 Memory Banks
Bank 0 : 256MB DDR-SDRAM 2.5-3-3-7CL 1CMD
Bank 2 : 256MB DDR-SDRAM 2.5-3-3-7CL 1CMD
Bank Interleave : 4-way
Speed : 2x 170MHz (340MHz data rate)
Multiplier : 5/4x
Width : 64-bit
Power Save Mode : No
Fixed Hole Present : No

--end--
Win 2k All updates
MSI KT3 Ultra 2 Pro
512MB DDR Memory
ATI 8500 /128 4X Graphics
AC97 Audio from VIA chipset (really noisy, but functional 5.1 Chan audio)
IBM Deskstar ATA 100 8MB buffer IDE
Latest Bios updates.

WinUAE is most certainly SLOWER on the ATHLON than the 2.0 CELERON p4 at work. I now think it's actually more than 112%.
I will compare a few Benchmarks with work tomorow on math etc using Sandra.

No virus killers or background intensive tasks are running on my system, I actually went in to task manager and made sure of that, I also turned off every unused system resource in the computer management section. it made no difference. Not that it should matter that much as I did noe of this on the XP Celeron at work and it has background tasks running all the time for network updates and it still kicks my 2400+  out to pasture.

If it is windows XP making this HUGE difference, it's one hell of an upgrade for the windows OS and I'm going to upgrade all my M$ PCs to XP. I seriously doubt that this is the case.

As for optimising code on the JIT side of things, I would think it would be easier to program optimised x86(global) instructions for JIT as opposed to writing for the P4 only.
you would know it would work on all Intel chips like Pentium 3 and below which have different execution techniques to the P4. Long Vs Short word instruction timing, needs to be done in a certain order or something like that. all changed with p4, which broke a lot of normal code. i think they call it sse2 coding or something.

[SHADES]

@ all

Ok, set Cas to 2.0 not that i expected any real differences.

Re ran AIBB benchmark
EmuTest
Slower result than a 040 CPU. NASTY

This is on my Athlon 2400

The XP celeron benchmark pulls the 040 off the screen. There's something really wrong here.

As for Celeron CPU being differnt, yes, just as the Athlon is different to the Intel. i used p4 as the spec as it's the arcitecture of the latest celerons with some serious limitations like the hyperthreading gone and less cache.
Still, the Athlon should be able to beat the pants off the celeron with out a doubt.

Benchmarks on XP celeron box tomorrow.

[SHADES]

@ All
OK people!
The information is in for those of you that are interested.

I used the exact same SiSoft Sandra for both PCs.
My 1st post will be the Athlon benchmarks and the second will be the celeron if you want to save and compare the two.

My test results show the Athlon beats the Celeron is just about everything. So now, I must think that the JIT or something is coded for P4 type CPUs only.

[SHADES]

@ ALL
ATHLON BENCHMARK
MEMORY Athlon
--Start--
Benchmark Results
RAM Bandwidth Int Buffered aEMMX/aSSE : 2030MB/s
RAM Bandwidth Float Buffered aEMMX/aSSE : 1935MB/s

Int Buffered aEMMX/aSSE (Integer STREAM) Results Breakdown
Assignment : 2036MB/s
Scaling : 2034MB/s
Addition : 2025MB/s
Triad : 2027MB/s
Data Item Size : 8-bytes
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 93% (estimated)

Float Buffered aEMMX/aSSE (Float STREAM) Results Breakdown
Assignment : 2039MB/s
Scaling : 1957MB/s
Addition : 1885MB/s
Triad : 1860MB/s
Data Item Size : 8-bytes
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 89% (estimated)

Performance Test Status
Run ID : ADS on Thursday, 17 February 2005 at 11:13:42 PM
Memory Used by Test : 256MB
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Page Size : 4kB
Use Large Memory Pages : No

Chipset 1
Model : VIA Technologies Inc VT8366/A,VT8367 Apollo KT266/A,KT333 CPU to PCI Bridge
Front Side Bus Speed : 2x 136MHz (272MHz data rate)
Width : 64-bit
Maximum Bus Bandwidth : 2176MB/s (estimated)

Logical/Chipset 1 Memory Banks
Bank 0 : 256MB DDR-SDRAM 2.5-3-3-7CL 1CMD
Bank 2 : 256MB DDR-SDRAM 2.5-3-3-7CL 1CMD
Bank Interleave : 4-way
Speed : 2x 170MHz (340MHz data rate)
Width : 64-bit
Maximum Memory Bus Bandwidth : 2720MB/s (estimated)

Features
MMX Technology : Yes
SSE Technology : Yes
SSE2 Technology : No
SSE3 Technology : No
EMMX - Extended MMX Technology : Yes
HTT - Hyper-Threading Technology : No
--End--

MATH Athlon
--Start--
Benchmark Results
Dhrystone ALU : 7738MIPS
Whetstone FPU : 3219MFLOPS

Performance Test Status
Run ID : ADS on Thursday, 17 February 2005 at 11:17:35 PM
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Number of Runs : 64000 / 640

Processor
Model : AMD Athlon(tm) XP 2400+
Speed : 2.05GHz
Model Number : 2400 (estimated)
Performance Rating : PR2965 (estimated)
Type : Standard
L2 On-board Cache : 256kB ECC Synchronous Write-Back (16-way, 64 byte line size)

Chipset 1
Model : VIA Technologies Inc VT8366/A,VT8367 Apollo KT266/A,KT333 CPU to PCI Bridge
Front Side Bus Speed : 2x 136MHz (272MHz data rate)

Features
SSE2 Technology : No
SSE3 Technology : No
HTT - Hyper-Threading Technology : No
--END--

Multi Media Athlon
--Start--
Benchmark Results
Integer aEMMX/aSSE : 19288it/s
Float aSSE : 20585it/s

Performance Test Status
Run ID : ADS on Thursday, 17 February 2005 at 11:18:54 PM
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Rendered Image Size : 640x480

Processor
Model : AMD Athlon(tm) XP 2400+
Speed : 2.05GHz
Model Number : 2400 (estimated)
Performance Rating : PR2965 (estimated)
Type : Standard
L2 On-board Cache : 256kB ECC Synchronous Write-Back (16-way, 64 byte line size)

Chipset 1
Model : VIA Technologies Inc VT8366/A,VT8367 Apollo KT266/A,KT333 CPU to PCI Bridge
Front Side Bus Speed : 2x 136MHz (272MHz data rate)

Features
MMX Technology : Yes
SSE Technology : Yes
SSE2 Technology : No
SSE3 Technology : No
EMMX - Extended MMX Technology : Yes
3DNow! Technology : Yes
Extended 3DNow! Technology : Yes
HTT - Hyper-Threading Technology : No
--End--

Cache/Mem/chipset
--Start--
Benchmark Results
Combined Index : 3130MB/s
Speed Factor : 23.5
2kB Blocks : 16750MB/s
4kB Blocks : 17666MB/s
8kB Blocks : 18155MB/s
16kB Blocks : 15285MB/s
32kB Blocks : 12690MB/s
64kB Blocks : 11823MB/s
128kB Blocks : 7819MB/s
256kB Blocks : 6453MB/s
512kB Blocks : 834MB/s
1MB Blocks : 775MB/s
4MB Blocks : 773MB/s
16MB Blocks : 772MB/s
64MB Blocks : 772MB/s
256MB Blocks : 771MB/s

Float SSE Cache/Memory Results Breakdown
Data Item Size : 16-bytes
Buffering Used : No
Offset Displacement Used : Yes

Performance Test Status
Run ID : ADS on Thursday, 17 February 2005 at 11:20:10 PM
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Page Size : 4kB
Use Large Memory Pages : No

Processor
Model : AMD Athlon(tm) XP 2400+
Speed : 2.05GHz
Model Number : 2400 (estimated)
Performance Rating : PR2965 (estimated)
Type : Standard
Internal Data Cache : 64kB Synchronous Write-Back (2-way, 64 byte line size)
L2 On-board Cache : 256kB ECC Synchronous Write-Back (16-way, 64 byte line size)

Features
SSE Technology : Yes
SSE2 Technology : No
SSE3 Technology : No
EMMX - Extended MMX Technology : Yes
HTT - Hyper-Threading Technology : No

Chipset 1
Model : VIA Technologies Inc VT8366/A,VT8367 Apollo KT266/A,KT333 CPU to PCI Bridge
Front Side Bus Speed : 2x 136MHz (272MHz data rate)
Width : 64-bit
Maximum Bus Bandwidth : 2176MB/s (estimated)

Logical/Chipset 1 Memory Banks
Bank 0 : 256MB DDR-SDRAM 2.5-3-3-7CL 1CMD
Bank 2 : 256MB DDR-SDRAM 2.5-3-3-7CL 1CMD
Bank Interleave : 4-way
Speed : 2x 170MHz (340MHz data rate)
Width : 64-bit
Maximum Memory Bus Bandwidth : 2720MB/s (estimated)
--END--

[SHADES]

@ ALL
CELERON BENCHMARK

MEMORY CELERON
--Start--
Benchmark Results
RAM Bandwidth Int Buffered iSSE2 : 1954MB/s
RAM Bandwidth Float Buffered iSSE2 : 2325MB/s

Int Buffered iSSE2 (Integer STREAM) Results Breakdown
Assignment : 2093MB/s
Scaling : 2103MB/s
Addition : 1816MB/s
Triad : 1807MB/s
Data Item Size : 16-bytes
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 74% (estimated)

Float Buffered iSSE2 (Float STREAM) Results Breakdown
Assignment : 2297MB/s
Scaling : 2286MB/s
Addition : 2366MB/s
Triad : 2354MB/s
Data Item Size : 16-bytes
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 88% (estimated)

Performance Test Status
Run ID : H000AUB42304490 on Friday, 18 February 2005 at 7:53:42 AM
Memory Used by Test : 252MB
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Page Size : 4kB
Use Large Memory Pages : No

Chipset 1
Model : Hewlett-Packard Company 82865G/PE/P, 82848P DRAM Controller / Host-Hub Interface
Front Side Bus Speed : 4x 100MHz (400MHz data rate)
Width : 64-bit
Maximum Bus Bandwidth : 3200MB/s (estimated)

Logical/Chipset 1 Memory Banks
Bank 0 : 512MB DDR-SDRAM 2.0-3-3-6CL 1CMD
Shared Memory : 8MB
Channels : 1
Speed : 2x 166MHz (332MHz data rate)
Width : 64-bit
Performance Acceleration Technology : Yes
Maximum Memory Bus Bandwidth : 2656MB/s (estimated)

Features
MMX Technology : Yes
SSE Technology : Yes
SSE2 Technology : Yes
SSE3 Technology : No
HTT - Hyper-Threading Technology : No
--END--

MATH CELERON
--Start--
Benchmark Results
Dhrystone ALU : 6109MIPS
Whetstone FPU : 1739MFLOPS
Whetstone iSSE2 : 3206MFLOPS

Performance Test Status
Run ID : H000AUB42304490 on Friday, 18 February 2005 at 7:51:28 AM
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Number of Runs : 64000 / 640

Processor
Model : Intel(R) Celeron(R) CPU 2.40GHz
Speed : 2.39GHz
Performance Rating : PR2633 (estimated)
Type : Standard
L2 On-board Cache : 128kB ECC Synchronous ATC (2-way sectored, 64 byte line size)

Chipset 1
Model : Hewlett-Packard Company 82865G/PE/P, 82848P DRAM Controller / Host-Hub Interface
Front Side Bus Speed : 4x 100MHz (400MHz data rate)

Features
SSE2 Technology : Yes
SSE3 Technology : No
HTT - Hyper-Threading Technology : No
--END--

Multimedia CELERON
--Start--
Benchmark Results
Integer iSSE2 : 14765it/s
Float iSSE2 : 18500it/s

Performance Test Status
Run ID : H000AUB42304490 on Friday, 18 February 2005 at 8:07:04 AM
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Rendered Image Size : 640x480

Processor
Model : Intel(R) Celeron(R) CPU 2.40GHz
Speed : 2.39GHz
Performance Rating : PR2633 (estimated)
Type : Standard
L2 On-board Cache : 128kB ECC Synchronous ATC (2-way sectored, 64 byte line size)

Chipset 1
Model : Hewlett-Packard Company 82865G/PE/P, 82848P DRAM Controller / Host-Hub Interface
Front Side Bus Speed : 4x 100MHz (400MHz data rate)

Features
MMX Technology : Yes
SSE Technology : Yes
SSE2 Technology : Yes
SSE3 Technology : No
HTT - Hyper-Threading Technology : No
--END--

Cache/Mem/Chipset CELERON
--Start--
Benchmark Results
Combined Index : 4497MB/s
Speed Factor : 18.8
2kB Blocks : 25460MB/s
4kB Blocks : 25998MB/s
8kB Blocks : 21072MB/s
16kB Blocks : 18017MB/s
32kB Blocks : 18628MB/s
64kB Blocks : 18624MB/s
128kB Blocks : 14275MB/s
256kB Blocks : 1604MB/s
512kB Blocks : 1381MB/s
1MB Blocks : 1390MB/s
4MB Blocks : 1398MB/s
16MB Blocks : 1402MB/s
64MB Blocks : 1399MB/s
256MB Blocks : 1395MB/s

Float SSE2 Cache/Memory Results Breakdown
Data Item Size : 16-bytes
Buffering Used : No
Offset Displacement Used : Yes

Performance Test Status
Run ID : H000AUB42304490 on Friday, 18 February 2005 at 8:09:11 AM
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : No
Page Size : 4kB
Use Large Memory Pages : No

Processor
Model : Intel(R) Celeron(R) CPU 2.40GHz
Speed : 2.39GHz
Performance Rating : PR2633 (estimated)
Type : Standard
Internal Data Cache : 8kB Synchronous Write-Thru (4-way sectored, 64 byte line size)
L2 On-board Cache : 128kB ECC Synchronous ATC (2-way sectored, 64 byte line size)

Features
SSE Technology : Yes
SSE2 Technology : Yes
SSE3 Technology : No
HTT - Hyper-Threading Technology : No

Chipset 1
Model : Hewlett-Packard Company 82865G/PE/P, 82848P DRAM Controller / Host-Hub Interface
Front Side Bus Speed : 4x 100MHz (400MHz data rate)
Width : 64-bit
Maximum Bus Bandwidth : 3200MB/s (estimated)

Logical/Chipset 1 Memory Banks
Bank 0 : 512MB DDR-SDRAM 2.0-3-3-6CL 1CMD
Shared Memory : 8MB
Channels : 1
Speed : 2x 166MHz (332MHz data rate)
Width : 64-bit
Performance Acceleration Technology : Yes
Maximum Memory Bus Bandwidth : 2656MB/s (estimated)
--END--

[SHADES]

@ Bloodline

I'm sure I already have said that I am using exactly the same WinUAE settings from work at home.
I copied the entire WinUAE directory from the work PC and installed it as is on my home PC just to be sure!

It most certainly can not be WinUAE settings however you may be right with using that term settings as it could be my Windows settings. I have no idea if Windows XP is superior to Windows 2000 so you may be on to something.

The funny thing is, it's just such a HUGE difference in performance. It's almost like I have 4 CPUs working together in the work PC when I compare it to the home PC.

I wouldn't think that windows XP could make that kind of an improvement.

Well that settles it then, there seems to be only one thing left to do now.
Install Windows XP on the home PC and test it all over again.
I have to get a copy of the XP OS first, I never liked XP so I only ever bought 2000. I don't like being told what I can or can't upgrade on my computer and Windows XP has that protection to stop you upgrading key components like CPU etc.  Wrong wrong wrong...Must kill M$

@ all
What do you all think??

[SHADES]

@ D

Hi, ALL WinUAE settings are identicle for both PCs.
It is a direct copy of the work PCs WinUAE directory just to remove any doubt on my end.

I can't overclock the FSB too much, my tollerence for other components don't like to go too far and light overclocking on memory etc didn't show much improvement on AIBB benchmarks.
I agree in the software problem, just trying to find where that could be in Windows is beyond my knowledge. Windows is a pig.

[SHADES]

@ D

Hmm.
I'll make some tonight, LHA em up and mail them to you if you like?

It will have to be when I get home as I'm at work now.

[SHADES]

@ D

True, they would differ under emulation. Most things will as the JIT would cause havoc.

Even so, it dosen't explain why the work PC feels so much faster graphicaly than the PC at home, it's what made me try a few benchmarks in the first place, just to make sure I wasn't just feeling things.

In AIBB I run the default PAL 8 colour tests. No UAE native modes, just classic AMIGA modes for the testing. I do the same  for both PCs.
Really, it's not going to be accurate however, it also should't deviate to the point of being unrealistic from similar speed PCs.
It gave me the opinion that something funny was indeed going on and that the PC at home (win2k PC) was most certainly seeming slower than the XP PC at work. That has lead me to this point. Testing with the help of others.

I'm not sure what's going on here, but the Celeron XP PC certainly  seems to be faster than my Athlon, and the strange thing is, the Athlon is the faster PC.

[SHADES]

@ ALL

Ok, this is really strange, but because i don't have a windows XP install disc, I resorted to re-installing windows 2000 from scratch.

After that and installing all the relevent patches, i re-installed WinUAE from the archive I had from the work PC and it is now functioning faster.

I'm not sure what was wrong in my windows setup, but there you have it, problem solved.

It just goes to show you how BAD windows really is and it's just another reason I now have to hope Hyperion hurry up a little and bring out OS4 because I no longer want to run this garbage.

Thanks to you all for your help here, it certainly was interesting to find out this fault.