PDA

View Full Version : New Shit


Asher
24-09-2003, 07:07:22
Athlon 64 3200+
Sapphire Radeon 9800 Pro 256MB
Zalman noiseless video card cooler
Zalman noiseless "flower" CPU cooler
Antec Sonata ultra-quiet case with noiseless powersupply
1GB DDR400, CL2
ASUS nForce3 150
Audigy 2

ETA: Friday

Sir Penguin
24-09-2003, 08:00:26
Let us know how that works out for you.

SP

Funkodrom
24-09-2003, 09:09:16
Yep, that's shit.

Qaj the Fuzzy Love Worm
24-09-2003, 17:16:21
Won't you need a monitor with that? :)

No longer Trippin
24-09-2003, 19:47:22
No hard drive(s)? What your not waiting for the rebranded Xeon, P4EE, it's only set to be 800 dollars on launch? :)

fp
24-09-2003, 21:03:41
He's going for two 120GB hard drives linked with a fancy RAID mobo.

Asher
25-09-2003, 05:01:50
Originally posted by No longer Trippin
No hard drive(s)? What your not waiting for the rebranded Xeon, P4EE, it's only set to be 800 dollars on launch? :)
Want instant satisfaction.

And I forgot to list HDs, 2 x 120GB 7200rpm 8MB cache SATA, in a RAID0.

Asher
25-09-2003, 05:04:39
Originally posted by Qaj the Fuzzy Love Worm
Won't you need a monitor with that? :)
I've got two NEC MultiSync FE950+'s laying around that work just fine. 19" Trinitron CRT.

Sir Penguin
25-09-2003, 06:02:34
240 GB? There can't be that much porn in the world. :clueless

SP

No longer Trippin
25-09-2003, 15:15:11
Why RAID 0 what is essentially an IDE drive with IDE tolerances. I've heard too many horror stories of one drive going down taking out the array. Not a fun thing for me at least having to reinstall everything. I'd opt for 1 or 0+1 so you'd get the redundancy.

Asher
25-09-2003, 18:15:17
RAID0 doubles the reading and writing speeds.

No longer Trippin
25-09-2003, 21:23:28
Sorry to burst your bubble here on RAID 0, but your going on theory, in reality it is MUCH different. Theory doesn't take into account disk allocation, access times, and several other factors, it only goes by what it can do in feeding instructions, and that doesn't mean squat if the drive is still sitting on a request as it's accessing it, thus it waits until a drive opens, then it sends the request. SCSI is the only drives you see a real benefit from RAID.

Like I said, double the speed, theoretically. On IDE ATA and most current SATA only gives you about 10 to 20% more speed respectively due to how insctructions are handled and true bandwidth. SCSI handles 256 instructions per request independently, IDE ATA and most (if not all) SATA handles one instruction per request independently. Once you average out all the R/W/access benches you get a whopping 10% boost for double the failure rate, you might hit 20% with SATA due to it actually being able to use it's marked transfer speed at least on burst of (SATA) 150 - ATA133 doesn't approach that, it was a marketting gimmick, hence why WD's IDE ATA100's still run faster than an Maxtor running ATA133. It can't run at that speed period other than for testing purposes in which it was certified in all actuality - it only became standard as a marketting push and it wasn't truly looked into to see if it could deliver the speeds as promised, so far ATA100 is as good as IDE gets. If you want to get anything out of RAID you need SCSI for the enhanced instructions per request. 256 compared to 1 makes a huge difference once in a RAID configuration - not to mention the added quality which would have me think of sticking it on RAID 0. The only SATA drive I'd do that with right now is the WD Raptor series as they are built to SCSI tolerances, the others aren't and the warranty shows it. 10-20% MAX gain for a surefire double failure rate. I'd suggest backing up quite often as the 120's are built to IDE tolerances.

One source with some good explanations and such: www.storagereview.com

Drekkus
25-09-2003, 21:29:47
Originally posted by Asher
Athlon 64 3200+
Sapphire Radeon 9800 Pro 256MB
Zalman noiseless video card cooler
Zalman noiseless "flower" CPU cooler
Antec Sonata ultra-quiet case with noiseless powersupply
1GB DDR400, CL2
ASUS nForce3 150
Audigy 2

ETA: Friday :eek: Holy fuck, that's nice. Did your parents neglect you as a kid??

No longer Trippin
25-09-2003, 22:03:14
I don't think so, his parents gave him a notebook worth a mint not to long ago - though he was going to pay for it himself to his credit, just they paid. Can't argue with him there. If I had to build a new system and my current system was old, I'd probably have went with a similar setup since I can afford to put money in AMD's pockets now - and they need that. Just I'd opt for the FX51 even though it needs registered DIMMs (thus slower memory) and cost more - though is faster. Also built on the Opteron which is why it needs that. By 6 months or so it will not need the registered dimms (FX53 most likely). Though personally since I can run my barton at speeds which equal to surpassing a 3.2 C (a bit over 2.4 ghz, not the BS PR marking) I see no need in jumping to a whole new board and chip especially since vid cards are soon to go PCI-X and and that isn't reverse compatible with AGP. Add in BTX form factor change and DDR 2 along with some huge SATA leaps on the horizon, I'd be wasting my money for me, though for him, he isn't since he wants what is nearly as fast as possible without overclocking. From what I remember of Asher's old machine though, he was due for an upgrade IMO (but I like overkill :) ).

Asher
26-09-2003, 05:28:27
Thanks for the input, but I just did a project on RAID for CPSC461 and I know how it works. :)

I think the information you're working off is a bit outdated. Check this out:
http://www.vr-zone.com/reviews/Seagate/BarracudaV/hdtach-randomaccess.gif
0.3ms slower access time (negligable)

http://www.vr-zone.com/reviews/Seagate/BarracudaV/hdtach-maxreadspeed.gif
http://www.vr-zone.com/reviews/Seagate/BarracudaV/sandra2003-filesystem.gif
Speaks for itself :)

Asher
26-09-2003, 05:31:52
Originally posted by Drekkus
:eek: Holy fuck, that's nice. Did your parents neglect you as a kid??
I'm buying this one all on my own, the 'rents coughed up the cash for my ThinkPad T40 because that was for school. And I was going to buy that one too, they just kinda said they'd pay for it and not to argue, since they bought my brother a ton of stuff for college this year too.

Asher
26-09-2003, 05:36:07
For the record, this rig only set me back $2500CDN, or $1850US / 1600 euro / 1100 pounds

No longer Trippin
26-09-2003, 19:38:25
That's about what I would expect, the largest cost being the CPU and vid card.

Random Access times is generally the one that will be what drops your speeds in real usage - and you got only a marginally better score. Burst speeds are good for only a few times after a defrag in some instances, in others they last a good bit longer - though you start taking more and more of a hit after a certain point that it's ridiculous at the speeds you get (generally on nearly full drives). Loading a game or a level, you'll be pulling burst for the most part, when you get to saves and such, you get more random access times. Then you have your general fragmentation over time. That was what I was talking about and your benches showed it. I knew your burst times would be good, and that's what the benches showed there. Just that the Random times would be only marginally better at best. Though it's much cheaper than SCSI, and not much more than IDE at all if not the same price now.

Only thing I would do is have them as 0+1 instead of 0 due to failure rates like I've said before. Although I've yet to have a Seagate die on me (though they are all SCSI) - but still, recovering isn't fun.

Asher
26-09-2003, 21:19:14
The documentation on my motherboard says 0+1 will require 4 drives.

No longer Trippin
26-09-2003, 22:28:49
Ah, that sucks - and with 0+1 you need matching HDD's, that would add a few hundred to the cost. Not really worth it. I'd throw in an old HDD for copies of critical files just incase it goes down since it would be intact since it's off the array. Just pull a drive from your old comp if you can. I know you don't want to lose something for school that you need and be screwed.

Asher
01-10-2003, 06:20:54
The rig has been fully assembled, tested, and burned-in.

The thing is fucking totally noiseless. And VERY fast: I installed ALL of Office 2003 Professional Enterprise edition (5 CDs!) in 10 minutes. I imagine RAID 0 has a lot to thank for this.

The Zalman coolers, quite simply, kick ass.
Take the VGA Cooler heatpipe on the Radeon 9800 Pro for example:

WITHOUT artifacts or instability, I can hit 486MHz core (380MHz is the stock speed) and 420MHz memory (up from 680MHz). That's just fucking incredible, even the new Radeon 9800 XTs are only something like 420MHz core.

The CPU hits 2140MHz (up from 2000MHz) while idling at 33C and peaking at 40C under repeated demo runs of AquaMark 3.0 and 3DMark2003.

Asher
01-10-2003, 06:29:27
And I think the reason the 9800 Pro core couldn't get any faster was simply signal propagation in the chip, not heat: Video cards lock when they overheat. The 9800 Pro never locked, it just had white specs over the screen.

To me, that says some of the pipelines or shaders couldn't keep up with the rest of the chip and thus the white blocks.

No longer Trippin
01-10-2003, 10:05:50
Asher, sounds about right for what you should get overclocked without any vmodding - what nanosecond chips you have? I'd imagine you'd have gotten more than 420 unless it's the smasung 3.3ns memory. Common in vid cards, use the slowest memory possible when adding memory (was common with nvidia at least for the G4's - 64's had faster ram than 128's). Makes a ton of sense there when those who want more memory get it for performance. :rolleyes: Can't speak of ATI's of that time as I owned an nvidia until the 9800 Pro I have now.

To me I think you have memory artifacting. As that is the culprit 95% of the time for artifacts. When the core can't complete operations it generally just locks up or craps out completely until a reset. Try dropping your memory spein worse cases, just dies. If it still artifacts at stock memory, well then drop your core a bit till it stops, then raise the memory up in small increments. If it is your core, well drop it down a bit.

Asher
01-10-2003, 16:44:46
That's precisely what I did. I overclocked the core first til it started artifacting, then overclocked the memory as far as it'd go before artifacting.

The memory didn't do much, but it's on par with other 9800 Pros from what I've read online. The core was cool though. :)

No longer Trippin
01-10-2003, 18:45:37
What nanosecond ram - if it is 3.0 or 3.3 your not going to get much - those are the "slowest" speeds for a Pro's bios to accept the ram. Some are in the mid to lower 2's and they get great overclocks - just haven't seen any built that way lately, though it is in reference specs for acceptance. Just why increase cost when you don't have to because your competition sucks so much?

Well your getting memory artifacts most likely, it isn't starved for bandwidth as the gpu was never a memory hog compared to nvidia's lineup anyhow. That is why the G FX's get decent boost from the added 128 on top the traditional 128, they need it as so much is stored instead of rendered on the fly.

Deacon
01-10-2003, 21:35:21
That's a tripped-out system. I wanna Sonata too. :)

No longer Trippin
08-10-2003, 19:40:28
Oh, as for what I was talking about for performance on raid 0, here is the link that may make you think twice about it - it uses SR High End, Office, and Gaming benches, those are the ones which really matter - not burst rates, though it shows them as well. Not outdated considering the drives used.

http://www.storagereview.com/php/tiki/tiki-index.php?page=SingleDriveVsRaid0&PHPSESSID=87f1173f86615e7873513a33598babcd

Asher
09-10-2003, 02:37:21
I've been using a RAID 0 system now for a while and I simply can't go back.

Things install incredibly fast, load incredibly fast, and boots incredibly fast. :)

And I only see performance increases across the board on your link...

No longer Trippin
09-10-2003, 05:57:41
Err, read the article... the main benches show the 10% over what you'd get otherwise. The speed your seeing is the instruction set your drive uses. It is the same as SCSI, 256 compared to 1 on IDE. Your getting exactly what I'd expect, fast install and boot times, as well as load times for games, level times are the same though as one drive as it's unsequential, starting the game generally is one or two large files and a few smaller ones, thus that is where you get the boost there since it stays mostly sequential unless fragmentation takes place.

SR Drivemark shows unsequential reads/writes which is the most common access. Those increases are 10% or less. Drive access time actualy is a tad slower as well - so I wouldn't say across the board - the only gains above 10% are on the sequetial R/Ws. Installation of stuff will fly by as it is sequential. Loading of single files will be fast as well if they are large and unfragmented. Like game load times on startup for instance, but actual level loads and stuff of that nature would be nearly the same as an unRAIDed setup. That goes for any RAID setup. I get great boot, install, and initial startup of games or other large filed programs - but going by one drive, it's a bit faster, but not much more with the unsequential stuff.

Glad you like it - just get into a habit of backing up, seriously. I do it at least once a week and I have redundancy. I know several people who've gone RAID 0 to only bitch after a drive failure - one of which was a SCSI setup surprisingly. Atlas drives IIRC. I'm sure you have a DVD burner with what you spent, so backing up shouldn't be a problem.

--------

What's the overclocking options on the board like (I know your not an overclocker, but you know the architecture better than I as I haven't looked into the A64 or FX-51 much) - you still have multiplier unlocking and vcore voltage I assume? I heard the FSB is gone for clock speed increases - but is it still there along with vdimm? Now you adjust the hypertransport from what I've understood from what I've read but still murky on it. Still confused on that.

I went cheap since I don't need a new board and I just ordered a AXQEA Barton 2500 (my current 2500 is going into my "school" machine I'm building cheap so I don't lose anymore drives from creative installs of XP by the school from backups when transferring data) - those tend to hit 2.3 on stock, with 2.5 to 2.6 overclocks being average with a slight boost to the vcore. Don't really feel the need to splurge since it's just essentially a junk comp and my current rig will be running on par with a P4 EE hopefully, just can't blow the bridges to go higher due to the new laminate they have on them. But curiousity sparks an interest anyways as I'm going to migrate to the 64 FX line in a year or so.

Asher
09-10-2003, 06:43:06
I've nothing important on this computer, all of that stuff gets duplicated on my laptop if I do have anything, so I'm not worried.

For the Athlon 64s and 64FXs, the only thing preventing you from unlocking is the BIOS. Several hardware sites have access to the modified BIOSes which unlock the processors, but no one has leaked one yet as far as I know.

The HyperTransport is essentially the FSB. How I've overclocked it right now (from 2000 to 2140MHz) is by cranking up the hypertransport to 214MHz from 200MHz. The chip uses a 10x multiplier, so that gave me 2140MHz.

You can adjust pretty much everything on the Asus K8v: vcore voltage, AGP voltage, memory voltage, about a dozen different memory timings, and 1MHz increments for the HyperTransport bus from 200 to 300MHz.

Asher
09-10-2003, 06:45:14
AMD has a new driver for Athlon 64s for Windows XP that enables Cool n' Quiet. It's basically the new version of PowerNow, but in desktops. When the CPU is idle it drops down to 800MHz and slows down the fans.

Unfortunately, I have no "processor" tree inside Windows XP's device manager. And viewing AMD's new processor support forum, this is a very common problem, at least with the Asus K8Vs.

So until someone fixes that (MS, ASUS, or AMD) I can't update to the new processor driver.

No longer Trippin
09-10-2003, 15:29:41
Well glad you have some backup and don't have much to lose - that's my main concern with RAID 0 and anyone who uses it unless I absolutely hate them, then I won't warn them at all about the need for backups. If only the Tekstar series drives were profitable from a few years back you could have had 0+1 with 2 drives as it could do RAID 0 alone theoretically.

Probably won't be unlocking features for the chipsets for some time if ever. Not soon because they are still on batches with 40 to 50% yields and thus a lower tolerance to overclocking. I had heard something about hypertransport working in that sort of fashion, just didn't know exactly how. Only that hypertransport isn't just addressing your memory as the FSB would from what I've read. It's broader, thus covering all data flow from the chip.

Thanks for the heads up. I'm assuming by the downthrottling (something Intel is doing now as well) that it runs hot on a stock sink. Coolbits Silicon microtube technology shows a lot of promise if it can get past the R&D stage (micron sized heatpipes to a heatsink, thus eliminating hotspots. Carbon "paste" is showing promise as well, but you need a good bit of pressure for that to work, and no company is producing it even though it is dirt cheap to make. The San Diego should lower the temps a bit, but then they'll have a faster clock, so then that just brings them up. All the other features sound like common nforce2 features on the better boards.