Weird Stuff..

[Dave_07]

Hey all,

Just got the new system up, and have noticed some strange things. The system is as in sig.

1) First off, in the bios the "CPU Bus Clock(Mhz)" is set to 200 and the "DDR Clock(Mhz)" set to 400. Shouldn't a 800fsb cpu be 400 bus clock and then 800 ddr clock ?
Also there is abit in this section of the bios called Performance Mode, which is set to "slow" ... LoL this is very weird, you'd expect it to be set at High or Medium or somthing, the options in the perfromance mode section are: Slow, Fast, Turbo & Ultra Turbo, and it's set on slow, lol. This can't be right can it ?

2) Also the "DRAM Timing Settings" section is set up as follows
-------------------------------
Configure DRAM Timing by SPD = ENABLED

DRAM Cas# Latency 2 Clocks
DRAM Ras# Precharge 4 Clocks
DRAM Ras# to Cas# Delay 4 Clocks
DRAM Precharge Delay 6 Clocks
DRAM Burst Length 8

DRAM Integrity Mode = DISABLED
-------------------------------

Are these setting geting the most out of the Corasir 3200 (400mz) ram, actuatly I can't find any indication of what seed the ram is operating at anyway, so actuatly what im asking is are these timings what you'd expect to see for ram at 400Mhz in dual channel setup ??

3) The 2x80gb sata drives in sig are setup in raid at 0, but there isn't any mention of any 80gb drives or any raid arry's anywhere in the bios, lol. In fact the only time raid is even mentioned in the bios is in one line I think called Onboard Promise IDE, which is set to "As Raid" but this dosen't make any sense, as the drives are sata not ide which is a completly different interface. lol
Also in Device Manager in windows, it reads the drives as "Promise 2+0 Stripe/RAID0 SCSI Disk Drive" this is also off, as sata and SCSI drive's are different types of drives.

4) Lastly, there is another line in the bios called "APG Aperture Size" which is set to 64mb. Shouldn't this be 256mb as eg graphics card has 256mb onbaord, or is this aperture size thing nothing to do with the graphics card ?


Cheers for all help, Dave

[Ethos]

1) Cpu bus clock is quad pumped with the P4C(800fsb) so 200 x 4 = 800mhz fsb

400 "ddr clock" = pc3200 which is what it should run at on a default (not clocked) P4C.

2) Those settings are fine, you might be able to get em to run tighter though, try:

DRAM Cas# Latency 2 Clocks
DRAM Ras# Precharge 2 Clocks
DRAM Ras# to Cas# Delay 2 Clocks
DRAM Precharge Delay 5 Clocks

Your ram might not be able cope with such tight timings, but have a play about!

3) Have a look through ya manual for settings.... if your hard-disk set-up is performing well i wouldnt worry to much about what windows says.

4) Aperture size doesn't make any difference nowadays, set it to 128mb (why the hell not?!) and forget about it

Hope this helps, Ethos

[Austin]

Yup P4 use QDR, 200x4=800FSB while DDR is ... well ... DDR so 200x2=400mhz or DDR400-PC3200 in Dual Channel mode it's effectively equal to the FSB 2xDDR400. I'm not sure what perf mode relates to, maybe just o/c's the FSB a little or runs some timings very tightly (mobo, RAM etc), step it up a notch, benchmark and ensure you're stable ... if you are leave it set faster.

RAM timings set to SPD tends to be the safest setting and 2-4-4-6 isn't bad at all, esp on P4 which benefits less. Lower is obviously faster as there's less waiting. If you do tighten timings or o/c your RAM you need to ensure you are perfectly stable. IIRC Intel 800FSB mobos can behave a bit freaky (as in slower) when certain timings are used, benchmark to check the gains match up. I'd leave DRAM integrity mode disabled, it may be for ECC RAM but basically home users really have no need for it (not mission critical) and it will slow you down a bit too.

It's normal for SATA to be considered as SCSI, they are more SCSI like than the std ATA (PATA). AFAIK SATA is still IDE, just a more updated version than PATA with it's own channel and Serial instead of Parrallel. SATA is still early, I doubt many drivers, BIOS' and Windows itself will get it listed 'right'.

AGP Aperture Size is the MAXIMUM amount of system RAM which can be used by the gfx card when it runs out of its own gfx RAM. Even with 128MB that's very rare (read 3Dmark03 with 4xAA+8xAF and even then only on Nature), 256MB gfx RAM definitely nullifies it but some people believe it has hidden properties. Since the system RAM will only be utilised when the gfx card runs out there's no harm setting to anything up to your total amount of system RAM. One thing to note is that the Aperture Size is wasteful, 128MB Aperture using the maximum 128MB of system RAM only gives the gfx card an extra 58MB due to tables, virtual addressing etc. Those with 64MB cards really want to set it to 256MB, for cards with 128MB and certainly up it's virtually pointless at the mo.

[Dave_07]

Cheers Ethos, Austin.
So basically all is well, pheww, I was so expecting somthing to go wrong, I was half expecting the system to not even boot lol, but looks like that luck is with me this time.
oh, how do p4's benefit less from 2446 timings ?
Ps, In the my computer bit in windows, there are only two hard drives shown, now I know raided drives are seen as one drive, but the two drives shown have generic name's eg Local Disck (c) & New Volume (d) with xp installed on the Local Disck (c) drive, how are mean't to tell which is the two 80gb raided drives and the 160gb normal stand alone drive ? Both drives are shown as having the same capacity.

Cheers, Dave

[Austin]

The archy of P4 and SktA CPUs is very different, P4 much prefers raw mhz over timings while SktA gains very well from faster RAM timings. I guess it's easier to fulfill SktA's memory bandwidth demands mostly due to the lower FSB speed. Gains are there on the P4 but generally they're about half of the gains you get on SktA when applying the same faster timings. It seems the Intel 800FSB chipsets have a sweet spot for RAM timings, for some reason faster timings isn't always yielding faster perf ... may be something to do with the way it handles Dual Channel?

To determine which drive is the RAIDed one you can simply try benchmarking ... or simply unplug the single 160GB HD and see if Windows boots and what drives are then detected.

[TiG]

Sorry Austin how does the architecture differences have anything to do with memory timings?.

The system still has to wait for responses from memory to keep the pipeline going.

Sorry but I doubt if you've got enough knowledge of the instruction set of either processor to do anything other than have a stab in the dark at that sort of indepth knowledge.

If they appear as two seperate drives in windows they are not raided, windows should not be able to see two drives if they are properly raided, the raid config should be done in the seperate raid controller setup which is usuallly shown after bios setup and is usually accessed via Ctrl A or some other sequence of keys,

The post screen before windows should show you this.
You will then have to initialise the drives and set them into the correct raid config. I assume you probably want striped so i'd suggest using the standard strip size and just raid the whole disk. Then install windows and press f6 and install your raid controller diskette and find the correct drivers for your operating system.

You can then partition the raided drive as required for your usage. See the thread on partitioning from a day or two ago for some general concensus on best methods of doing this.

Hope this helps
TiG

[Austin]

The basic principles are the same but the internal way in which the Athlon and P4 go abot things is different just as we see in the Rad8500-9200 vs GF3/4TI and Rad9500-9800 vs GF-FX. The products are different in their design, internal structure and implimentation. We see the P4 requiring much higher FSB and memory bandwidth to perform well, not so for the Athlon. Just as we see the GF-FX generally require higher clocks, much higher memory bandwidth, optimised code etc. It's been proven on many occasions that the Athlon really benefits quite significantly from lower RAM latency while the P4 though still benefits shows smaller gains. Why should memory timings have anything to do with pipelines or instruction sets? I'm talking pure realworld perf which is what counts, not paper specs and theory. I hope that addresses your point.

I'm not certain if you grasp that Dave_07 has a total of three drives, 1x160GB and a RAID array with 2x80GB=160GB. It isn't (AFAIK) that his RAID drives are showing up as two seperate drives but that he is unsure which of the '160GB' HDs is his RAID setup.

[TiG]

Nope Austin it doesn't, becuase thats exactly where it matters, real world performance doesn't happen by accident. Its how the processor works out what to do.

I don't see what Graphics cards have to do with the discussion at hand, infact i only see it as a bullrubbishrubbishrubbishrubbish tactic, I don't have any problem with you Austin, I think your a nice guy and help people a lot.

I do have issues when the information you give is dubious and absolutely subjective at best.

The Cycles measured in memory relate to 1 processor cycle, this fits in critically with processor pipelines, a memory fetch delay causes big problems for processors and the longer the pipeline the worse the effect, this is why the P4 needs high mhz, bigger pipelines need this to keep the output high, especially if there is branches in execution.

I guess only Dave_07 can answer that question on the disks, as i've seen no evidence that he's been in the raid controller bios i'm assuming its not raided and not working.

sorry Dave_07 don't want to impact you getting your machine fixed.

Thanks
TiG

[Stoo]

yup, AMD processors traditionally do more instructions per clock cycle than Intel processors, hence why Intel needs silly clock speeds and QDR sillyness to keep things even.
Because Intel processors handle less instructions per clock, memory timings have less of an impact because more often than not it's actually the ram that's waiting for another clock cycle to do something with the data.
With AMD processors doing more instructions per clock, keeping the lag down to a minimum with tighter memory timings benefits the system a lot more because the processor can actually do something with the data from the ram

At least that's what I seem to remember from the computer hardware lectures at uni a while ago

[Austin]

What I'm trying to say TiG is that because the P4 and Athlon work in very different ways they respond differently to different optimisations such as higher FSB, memory bandwidth (eg Dual Channel), RAM timings etc. What is important is real world perf, we can talk internal designs etc until the cows come home but bottom line is that 2-2-2-5 vs 2.5-3-3-7 shows roughly 2-3% over all perf boost on P4 and 3-5% perf on AthlonXP. The reason is because they go about the same job in different ways. They conform to the same basic stds much like Rad8500-9200 and GF3/4TI conform to DX8 (likewise Rad9500-9800 and GF-FX with DX9) but they go about things very differently internally. Hence GF-FX shows big gains from extra memory bandwidth while the Radeon design doesn't need or use it, gains will be there but they will be far smaller.

That's why I bought gfx cards in to the discussion, they have many parrallels with P4 vs AthlonXP in terms of conforming to a given std and producing the same output form the same input but in different ways. Since it's a small side point anyway I don't see what good it really does to dig so deep, maybe start your own thread TiG?

[TiG]

Stoo what you are talking about is MIMD/SIMD/SISD. This also exists in the Intel system, but this is what i'm saying about the claim or this figure of about half the benefit. I'm saying that is completely subjective and there is NO way that you can say something like that, its complete stab in the dark and no way you can quantify this.

As with you this isn't googled knowledge this is something that i've been taught by a Professor and something that I feel have confidence speaking about.

Both intel and amd benefit from the tightest memory timings you can get away with. Thats the simple truth.

And what do you think QDR is if its not simd (m being 4 in this case)

And the most important thing that again affects real world performance more than anything is out of order branch instruction execution, means pipeline is complete rubbish and it gets scrapped, amd is better here due to the shorter pipeline.

The number of steps in AMD i don't know (think its 11 or 14 depending on the type of instruction being executed)
but intel is 20

But what you also have to take account of is number of instructions being run per second (or nano seconds as the case is)

20 * (1/3000000000) vs 14 * (1/2200000000) in effect for the first instruction, and that instruction has done absolutely nothing in real world sense.

I'm sure i could go on this forever and ever, but i'm sure that will not help Dave_07 get his computer sorted, which i think is the best way to move forward.

So please i suggest someone creates another thread elsewhere to continue this if necessary.

Thanks
TiG

[Stoo]

Tig: That's a very good idea, and I quite like this in depth stuff, it's having a rather pleasing effect of waking some neurons up

Dave: can you post a screenie of your computer devices listing at all?

[Dave_07]

Cheers for all the posts guys.
The 3 drives are operating and the raid arry is as well. I managed to get the sig of the raid to show it's self on post. It's just windows applying silly generic name's to annoy you
I've tried looking at all the usualy clues to differentiate between the

[Stoo]

Yeah, if you have a lot of drives and RAID it can get a bit confusing, one of the ways around it is to have a look at the device manager and change the view to "Devices by Connection", then you can track through which interface has which drives, and if you open up the drive properties in the device manager and select the volumes tab, there's a populate button that will show the windows volumes on a particular drive

Bit of a cludge, but it works

[Dave_07]

YaY! Your a genius Stoo, It worked, raided drive = c & Stand Alone drive = d
Brilliant. Now see if you can you solve this one: For some reason my system won't go on stand by. When I click stand by the monitor gose to stand by (Sleep mode) but the base unit just shut's down
Then when I restart the system the monitor won't display anything, so I have to turn of the monitor & base unit and then wait for abit then turn them all back on.

Cheers, Dave.

[Austin]

Sounds like something doesn't like sleep mode. Check BIOS, Windows settings and may be even consider drivers? If you really need sleep mode try running with only the minimum components and see if you can find a culprit. Otherwise take a trip to the M$ KB.