-

[THUGSROOK]

JTKGY
dont get yourself all crazy over this...

333DDR Cas2/2/2/6 (manually set) is just about as fast as 400DDR Cas2.5/4/4/7. (in gaming benchmarks)
sandra is about the only test that says otherwise.

i use prime95, 3dmark2k, 3dmark2k1, revolt, quake3, elite force, X-Isle demo, serious sam, serious sam2, scisoft sandra, memtest86, divx encoding, etc.

 

[oldfart]

<< for a quicker test i just play serious sam1 or 2 - those games are SERIOUSLY hard on a system and will crash very quickly if there is any instability. >>


I find that SS is a more enjoyable stress test than watching those silly numbers scroll by in Prime95!!

 

[Clevor]

"comparisons vs AMD is yet to be seen.
id also be interested in how well RDram would do at the same 2.4ghz overclock."

I currently have a P4T-E running 144 FSB/4X RDRAM, so that puts me at PC1152. I get a 3550/3500 memory bandwidth. But very few people can get the RDRAM this high at 4X, and it's only 95% stable (minor memory problems, but most 3D runs fine). What the RDRAM needs is a bit more voltage, but nobody knows how to do this yet (except some mad overclockers in Japan) and it would require soldering.

So I have a 1.6a up to 2.3 gig. With a Ti 4600 overclocked to 315/735, I get a 11,750 3DMark2001 score and a 15,550 3DMark2000 score. Remember the scores depend a lot on what video card you have. I haven't checked the MadOnion database, but I'm sure some people have done better with the same rig. But I am running the 27.42 Detonators, and I hear the latest 28.90s are good for several hundred points so that might take me over 12k.

I just picked up a used P4B266-E here in Japan, and I have some Samsung PC2700 DDR333 on the way. I want to see how high I can take the same CPU on the DDR rig.

I really doubt you can get 3000/3000 at just 150 fsb with DDR, despite what one guy posted. All the benches I've seen show maybe 2800/2700 at 180+ FSB or something. I shall see for myself. With RDRAM at 133 FSB or PC1066, you should see around 3300/3300.

Does more bandwidth really help? I ran 3DMark2001 at 138 FSB/4X (2.2 gig, PC1104) versus 144 FSB/3X (2.3 gig, PC864). So it's 3400/3400 bandwidth with slower CPU versus 3000/3000 bandwidth and faster CPU. I got 300 more 3DMarks with the higher bandwidth setup. Granted that is only a 0.1 gig difference in CPUs.

 

[JTKGY]

<< JTKGY
dont get yourself all crazy over this...

333DDR Cas2/2/2/6 (manually set) is just about as fast as 400DDR Cas2.5/4/4/7. (in gaming benchmarks)
sandra is about the only test that says otherwise.
dont worry about 400DDR (or any DDR) until its Cas2/2/2 stuff.
same old story....

i use 3dmark2k, 3dmark2k1, revolt, quake3, elite force, X-Isle demo, serious sam, serious sam2, scisoft sandra, memtest86, divx encoding, etc.

IMHO - looping 3dmark2k or the X-Isle Demo (uses serious sam engine) for 12hours is the toughest thing i could do to my system. if it passes those, i concider it 100% stable.
for a quicker test i just play serious sam1 or 2 - those games are SERIOUSLY hard on a system and will crash very quickly if there is any instability. >>




THANKS!
I feel much much better now...
As for the Serious Sam... where could I download a demo copy or something to do
system quick torture tests?

 

[oldfart]

<< really doubt you can get 3000/3000 at just 150 fsb with DDR, despite what one guy posted >>


That would be me.

Oh yes, you certainly can.

<<< Memory Bandwidth Benchmark >>>
------------------------------------------------------------------------------------------------

< Benchmark Results >
RAM Bandwidth Int Buffered iSSE2: 3037MB/s
RAM Bandwidth Float Buffered iSSE2: 3031MB/s

< Int Buffered iSSE2 (Integer STREAM) Results Breakdown >
Assignment: 3038MB/s
Scaling: 3038MB/s
Addition: 3037MB/s
Triad: 3036MB/s
Data Item Size: 4-bytes / 32-bits
Buffering Used: Yes
Bandwidth Efficiency: 95% (estimated)

< Float Buffered iSSE2 (Float STREAM) Results Breakdown >
Assignment: 3032MB/s
Scaling: 3033MB/s
Addition: 3029MB/s
Triad: 3031MB/s
Data Item Size: 8-bytes / 64-bits
Buffering Used: Yes
Bandwidth Efficiency: 95% (estimated)

< Test Statistics >
Array Memory Used: 256MB
SMP Test: No
Total Test Threads: 1
SMT Test: No
Dynamic MP/MT Load Balance: No
Processor Affinity: No

< System Chipset >
Model: Intel Corporation 82845 Brookdale CPU to I/O Bridge
Front Side Bus Speed: 4x 150MHz (600MHz data rate)
Max. Bus Bandwidth: 4800MB/s (estimated)

< Logical/Chipset Memory Banks >
Bank 0 Setting: 256MB ECC DDR-SDRAM
Bank 1 Setting: 256MB ECC DDR-SDRAM
Memory Bus Speed: 2x 200MHz (400MHz data rate)
Max. Memory Bandwidth: 3200MB/s (estimated)

< Instruction Set Support >
MMX Technology: Yes
SSE Technology: Yes
SSE2 Technology: Yes
Hyper-Threading Technology: No

 

[THUGSROOK]

JTKGY
ill be 100% honest with you - you need to play with all of those settings 1 at a time to see how it effects performance and stability. you have alot of odd options - most of them are for memory timings tho.
start on its default settings and attempt to speed them up. these tests could take you weeks, but thats the fun of OCing/tweeking settings that work for me (Asus P4S333) may not work for you. (Soyo SiS645)
ill be happy to answer any individual questions as you go along tho
3dmark2000 would be a good gaming benchmark to use for this, as it is very sensative to memory performance and crashes easily.
Serious Sam 1 and 2 demos are available on the net - just do a search.

oldfart
<<< I find that SS is a more enjoyable stress test than watching those silly numbers scroll by in Prime95!! >>>
my sentiments exactly!
yep, 3000/3000 is very obtainable - ive got those scores myself at certain settings/ratios.

Clevor
the P4B266 doesnt have any damn memory ratios after 132fsb! ...yet
i wish it did - i have one just sitting here.
gluck with your tests tho!

 

[Clevor]

Old Fart, I believe you are running 150 FSB, but a 3:4 memory ratio, so your ram is running 200 FSB. So it might be possible then, although most guys are getting lower than that.

What kind of Sandra arithmetic and multi-media CPU marks are you getting?

 

[oldfart]

<< Old Fart, I believe you are running 150 FSB, but a 3:4 memory ratio, so your ram is running 200 FSB. So it might be possible then, although most guys are getting lower than that. >>


Yes, that is how I am running. I just ran another test and got 3049/3046. Its the truth. Epox 4BDA i845 board rocks. I'll run the CPU benches.

 

[THUGSROOK]

Clevor
async memory ratios come complete each with their own latency issues.
3:4 and 4:5 ratios are alot easier to run then a 4:6 ratio.
the 4:6 ratio is a PITA!
i can run 400DDR in 3:5, or 4:5 mode all day long on 2.7v, but to run the 400DDR 4:6 ratio takes 2.9v and its only 99.9% stable.

...so i have no problem taking oldfart at his word since hes running a 3:4 ratio.
also i guy named Duvie runs his kingmax PC2700 memory @ 3:5 400DDR.

 

[oldfart]

SiSoft ALU, MM, Mem Benches

I run the DDR @ 2.6 V which is default on the Epox

 

[THUGSROOK]

ok oldfart now youre just plain braggin
im thinkin bout that epox . . . . . eeeeeeepox. hehe

but ive got a poor P4B266-C sittin here just waiting for a 3:4 ratio above 132fsb :frown:
if that ever happens, id be off this SiS board w/in minutes!

 

[oldfart]

It would be nice if Asus would release a BIOS that would do that. There are a lot of people that have the P4B266. The lack of that feature is the reason I got the EPOX (my first ever EPOX board) over the ASUS. I still see people recommending the P4B266 or MSI 845 whan asked about a P4 845 board. The EPOX is the one to get if you want the best performance from an 845 setup. The PCI/AGP 33/66 lock, 1.4 - 1.85 Vcore adj, boot from USB also are nice features. No cold/warm boot issues either. My one complaint is the active NB HS/fan. I replaced it right away with a passive Zalman.

THUGSROOK, I'd just sell the P4B and get the Epox. It's ~ $100 for the non RAID version, ~ $120 with RAID.

 

[THUGSROOK]

ever get the feeling that Intel "threats" are behind all the i845 limitations?

thats got to be the case - otherwise Asus wouldnt be such wimps about adding memory ratios to the bios!
we know they are capable, its not a chipset limitation.
maybe i just learn how to write my own bios files

:disgust:

 

[oldfart]

<< ever get the feeling that Intel "threats" are behind all the i845 limitations? >>


Good question. The 845 obviously can do the 3:4 @ any FSB. Does any other mfgr besides Epox enable the feature?

 

[jinks]

Hi.
The marking on Kingmax module is KDL 684T 4A2A-05.. Does that mean that it is -5. If that is the case then what memory module do I have?? Pc2100 or 2700.
Thanks
Jinx

 

[jinks]

Hi,
I checked but it seems that Bios Ver. 1006 is actually 1005 and 1007c is still a Beta version. Is that so??
Jinks

 

[THUGSROOK]

jinks you have 5ns PC2700 - prolly good for upto 400DDR speeds.
the P4s533 bios files are kinda screwy when it comes to file names - theres a note of this in the update section of the P4S533 bios thread.

 

[jinks]

Thanks Thugsrook for your prompt response. So Ihave 333DDR. Now how do I set it for maximum proformance let say 400DDR.
Jinks

 

[THUGSROOK]

just read thru the entire P4S533 bios thread in the "motherboards" forum and you should get a good idea of how memory ratios work. ive also posted a bunch of other threads in here that have info on the subject.

 

[THUGSROOK]

UPDATED - added cas 2/2/2 results

 

[THUGSROOK]

 

[jose]

Oldfart:

Have you tried your setup w/ 2 mem sticks installed ?

Soon I'll be getting a new mobo/cpu/mem combo, yours appears to be the best 845 solution.

I'm mainly waiting for RH7.3 to be released, because it has a newer install kernel that support the 845.

I need 1 Gig of memory, that's why I'd like to know if your system would still oc'd w/ 2 mem sticks (512mb).

Regards,
Jose

 

[THUGSROOK]

i only use 1 stick of memory.

 

[DoubleL]

I sure don't know how you are getting them 3DMark scores with the P4S533, I have one with PC-2700 memory a PNY GeForce 3 and the best it will do on 3DMark 2000 is 6800 running 2.133, Any Athlon I have here will blow that away

 

[THUGSROOK]

actually the motherboard is a P4S333 flashed with a P4S533 bios.
6000 sounds very very low for your system.