New hardware

[pascal]

I installed my new hardware this week.
The configuration now is:

- Asus P4B533 rev 1.04 BIOS 1015 (cheap, well tested, rock solid, probably most stable P4 mobo and ECC capable) *
- P4 2.4B retail (fast enough) *
- 3x256MB DDR 266 ECC Crucial (maximum data integrity for the workstation) *
- Asus V9560 Video suite (GFFX5600, good and cheap video combo) *
- SB Live
- HK speakers
- 3Com 10/100
- Logitech MX300 *
- 40GB Maxtor 7200 rpm
- TDK CDRW 16/10/40x
- generic midtower
- generic 400w psu
- Sony Trinitron
- WinXP Pro

* new iten

I changed from Athlon XP2100 to this P4 because I needed ECC for some tools I decided to have at home.

Anyway I tested the D3E3leak at 1024x768x32 with it and I disabled the VM´s page file (XP option). It is smooth and great 8)
I am impressed because it is using the nv20 path. The peak memory usage was 484mb. My problem now is that XP still need to acess the HD for some services, but no more VM´s disk swapping

Any ideas on how to improve it? Thanks

 

[chavvdarrr]

tools which require ECC memory to run ?! OMG you must be joking .... (will you PM the name of this software, just curious...)

as for improving:
2x512 MB ram
GF4 4200SE instead of 5600 ;P
bigger HDD

 

[Saem]

That's not that suprising.

If you're running anything which has data integrity as high priority it's a must.

It's likely some engineering or science app -- a simulator perhaps?

Regardless, have a look at the figures for chances of data corruption. Then do some quick math, you'll see with the technology of today, those figures aren't the very comforting.

 

[pascal]

tools which require ECC memory to run ?! OMG you must be joking .... (will you PM the name of this software, just curious...)

Cadence EDA tools. It doesnt require ECC to run, but I will feel safer. I have seen some memory corruption effects in other less capable hardware using this tools.

as for improving:
2x512 MB ram
GF4 4200SE instead of 5600 ;P
bigger HDD

I was thinking about improving the Win XP configuration.
More memory is always better.
The GF4 4200 will not be better with Doom 3 and I will miss the video suite.
Bigger HD will not make a difference for me.

Thanks for your reply

 

[chavvdarrr]

Cadence EDA tools. It doesnt require ECC to run, but I will feel safer. I have seen some memory corruption effects in other less capable hardware using this tools.
I was thinking about improving the Win XP configuration.
More memory is always better.
The GF4 4200 will not be better with Doom 3 and I will miss the video suite.
Bigger HD will not make a difference for me.
Thanks for your reply

Today's bigger HDD's, are not only bigger but faster too, p.e. newest disks from Maxtor, Seagate etc., use 80GB platters.
If you go using ECC, fearing of memory coruption, why not going RAID5 ? And if buying RAID controller & 3 HDDs may seem overkill, perhaps 2 HDDs in mirror will be safer? IMHO possibility of HDD failure is higher than possibility of ecc errors.

 

[pascal]

I disabled the VM´s page file then the HD performance impact is not that great.
Software RAID 1 (full disk mirroring) is something I could use to improve reliability/availability of the system. The relevant data I already keep out of the system.

Soft error corruption is something that happens constantlly. The classic IBM experiment show that a single DIMM will produce a fatal OS error every 750 hours. I have 3 DIMMs and then it means 3 times more possibility of fatal errors.

And before any fatal OS error many user data and system is corrupted. These tools sometimes show a memory acess error in some less capable hardware and I have personally experienced that.

Other day I saw a post of someone who compared the Windows installed in the system with the prototypes in the instalation CD. With ECC after a few months there was no diference, without ECC the OS was seriouslly corrupted. He now is very happy with Windows installed many years without a single problem using ECC.

If you are just gaming, net surfing or doing some small office stuff then dont worry, just install the OS from time to time. If you are working with very expensive data then start to worry a lot.

 

[chavvdarrr]

I disabled the VM´s page file then the HD performance impact is not that great.

Soft error corruption is something that happens constantlly. The classic IBM experiment show that a single DIMM will produce a fatal OS error every 750 hours. I have 3 DIMMs and then it means 3 times more possibility of fatal errors.

Other day I saw a post of someone who compared the Windows installed in the system with the prototypes in the instalation CD. With ECC after a few months there was no diference, without ECC the OS was seriouslly corrupted. He now is very happy with Windows installed many years without a single problem using ECC.

-I never advise disabling VM's page file - but I'm sure you know what you're doing.
- any link to that IBM experiment? 'cause in work we have machines running until shutdown for maintenance(~months), even without ECC
- without ECC the OS was corrupted? Well what about 40 iP100 running 5 years 8 hours a day Win95, without any corruption?

Anyway, wish you flawless hardware

 

[pascal]

Disabling the WinXP page file is no mystery, it is an option available in the OS. With lots of memory just drive carefully and dont try to open 50 IE or 50 words or play 2 UT2003 simultaneouslly

Serious, it is no problem if you know what you are doing and respect the memory limitations. Is like driving a manual gear car.

Links about IBM experiment changed but you can find more about ECC in the links below:
http://www.research.ibm.com/journal/rd/401/curtis.html
http://www.corsairdirect.com/ecc.html
http://www-1.ibm.com/servers/eserver/pseries/campaigns/chipkill.pdf

Yeah, the OS was (Windows NT or 2000, dont remenber) corrupted and it is a fact.

Well, soft errors really exist and are not imaginary and ECC exist to prevent it most cases. Servers and workstations use it. Is my option to be safe and use it
Thanks