Ought-8

New year, new blog entry, new... something. Re-new-all? Something like that. I haven't really been effective at using this space much in the past few months. There are a couple of reasons really - mainly I haven't really been enthusiastic about trying to use my computer. That and other toys have shown up at home (a DS Lite, the Wii from Santa) and the computer has been getting short shrift.

Soon should change. The biggest issue with my computer has been an electrical instability problem that really came to the fore after my original video card gave up on me. That card, a BFG6800GT-OC was an excellent card when I bought it. Plus it came with a lifetime warranty. So mid-2007, it's RAM died and it was replaced. The replacement wouldn't boot into WinXP properly, so I got a 7600GS to replace it. It worked well but the computer would suddenly shutdown in the middle of games. That is the electrical problem - when the CPU/GPU/Fans/HD draw a certain level of power, the motherboard can't take it and dies. I suspect that this was a manufacturing defect, given the quality of the parts and power supplies I used. I had problems from the first day setting up the machine, but the problems are really noticeable now. So I've ordered a new replacement AMD Socket 939 motherboard. Turns out there are very few choices of new 939 motherboards, with an AGP slot, left. This is an MSI with a VIA chipset. I'm hoping that the performance is similar but more stable.

The other thing I did to improve computer usage was to buy a nice LCD monitor during the "boxing week" period. Got a nice (21") but not cutting Samsung SyncMaster216BW. Spent less than half of the old 17" CRT it replaces, but certainly looks better. This thing looks much better than what I had. Gaming - I didn't notice the difference. Except it is brighter and sometimes the colours don't appear to converge properly in CS:S, but that's okay.

So hopefully the posts will be more frequent. It is a good thing to write a little every day. It would be good if that writing had a specific purpose or some other redeeming (monetary) quality to it, but one can't have everything.

Some answers (yah, right)

Took some time during lunch to call up BFG and talk to technical support about my computer lockup issue. The person I talked to listened and agreed when I suggested that both power supplies I have tried are good enough to support the card (7800GS). So the next thing to try would be to swap in a similar video card and to try this video card in a similar system. These things are not really possible (I try to have one copy of each generation of computer hardware, not more), so I asked him to speculate. He pointed out that the different answers would be the video card or the motherboard are the root causes.

At this point, I recalled the trouble I had with my ATX power connector when I first got the motherboard. I had trouble inserting the connector - I don't know if it was the motherboard or the connector itself, but something was causing issues. These problems were both mechanical and electrical - it was hard to plug in and sometimes no power flowed. So I think that is the answer. This is not really the answer I wanted as it means spending more money.

The first thing I thought to try is to see if I can find a replacement socket 939 motherboard with AGP. Such a beast does not really exist in the one place I checked (ncix.com), but it is possible that some place may have a new-but-old board in stock. And I'd be happy to relieve them of it. Otherwise, it means starting over.

Now, that is not necessarily a bad thing - it's been about 3 years and I could do with some minor upgrades. This is a good time to buy - Intel and AMD are in the midst of a price war, so good CPUs are plentiful. Ram is at a reasonable price and all the other parts except power supplies and video cards are nice and cheap. But because I don't have the funds at this time, I can't really go out and drop a bundle on that. Nothing is stopping me from speculating however...

I'd get either an Athlon 64 X2 6000+ (~$240) or an Intel Core 2 Q6600 (~$340). 2GB of RAM is about $120, a good case ~$150, a DVD burner ~$45, a decent HDD ~$120. Biased by these power issues, I'd get the power supply I have wanted for a long time, a PC Power & Cooling. This company was recommended by Jerry Pournelle for many years in Byte magazine. Every current review speaks very highly of their quality parts and design - always the most efficient, stable and very quiet. They were recently bought by OCZ, which basically means that they know are available where I can buy them. So I'd be wanting a beefy PC Power & Cooling Silencer 750. That's about $230, twice a "good" power supply and 2/3 more than an "ok" power supply. Finally, the tradeoff would be on the video card. A decent card can be had for ~$180, but you can spend $250, $400 or $800. If I had to do it right now, i'd get the $180 card because I don't have a spectacular monitor - anything modern will drive it well.

Anyway, for now I have to try and keep the video card underclocked - that seems to help prevent most of the power-related lockups. Not the best thing, but at least there is a workaround. The system is very responsive and stable otherwise, so I think this is a good compromise. I have to find a way to make the settings stick though.

The puzzlement continues

Didn't spend a whole lot of time on the computer today, except for this evening! I put everything through some excruciating trials - namely CounterStrike: Source. I was able to play for a few hours without any issues. But with a minor issue - I had to under-clock my GPU.

This is what I was doing from the beginning - that is when I had my original 6800GT OC. I had an "overclocked" card, but I had to underclock it to get it to work reliably. So I'm kinda torn - is it the power supply, the video card or something else unknown? Guess time will tell. I'm hoping that the computer store will take back their powersupply. I found that ncix.com has the Antec Neo HE 500 in stock, which looks like a good choice. It is a modular power supply with two of the output plugs being independent rails. Using that, I can put the hard drive, optical drive and fans on one rail, and the video card on another. With the Enermax Noisetaker II 495 (that's what the model is named), the manual seems to indicate that the second 12 V rail is for the CPU (that 4-pin plug right on the motherboard). That kinda defeats the purpose (for me anyway) of getting a split-rail powersupply.

Anyway, I'll also call BFG and see if they can diagnose what is going on. If they recommend a new powersupply, I'll see if they have a particular one in mind that I should use.

Quick post

So I thought I had everything figured out, but I didn't. Got a new power supply on Friday, an Enermax Noisetaker II 485W. It showed the same issue as before - while playing an intense game of CounterStrike: Source, the machine would hang. And we're not talking "blue screen" hang - hardware hang. Everything is stuck, but at least the fans keep turning.

Anyway, the only way around it is to reset and continue. I'm not sure what I should do now - I guess I will take back the power supply and get a larger one. That seems to be the next logical step. Otherwise, it will mean a complete upgrade. Not as if that is horrible, but it is something I really can't afford at this juncture. Gotta be impressed with Intel's latest price cuts though - they now have a quad-core chip for $266 (US). So it wouldn't be a horrible time to upgrade. But first a new power supply. I got a quote Friday for a 600W beast that would do the trick for sure, so maybe that is what I have to look at. Hopefully I can return this tomorrow (got it at Masters Computer). We'll see.

Further computation updates

As I mentioned here, I still haven't returned my OCZ ram to see if I can get it fixed. However, last Friday I did manage to send off my 2nd, not-properly-functioning BFG 6800GT. Today UPS showed up with the replacement. To my surprise and amazement, it was a BFG 7600 GS! Sweet! This card is a little newer and I was pretty sure that it used a little less power. Now that I have checked the specs more carefully, I think it withstands temperature better, is a smaller length, but has higher power requirements.

Specifically, when I compare the old power requirements with the new, the 6800GT requires a 300W power supply. I started off with a 430W power supply, but upgraded to an Antec TruePower 480 when I first put things together. Should be more than enough. The new card is more specific - 350W with a minimum 12V+ rating of 20 A. Those of you who keep with the vagaries of modern multi-video card systems will realize that this is pretty tamed compared to some Quad-SLI system and the 1100 and 1200W power supplies now available. When I was putting this machine together, the 6xxx series chips from NVidia represented the highest power consumption in video cards - the SLI and Crossfire technologies are only support on PCI-e. Such motherboards were available, but I went the AGP route, but I did buy a video card that required a lot of power. With an Antec 480 W supply, I thought I would be fine as Antec power supplies were part of a test group that reliably matched their number with their real output. What I didn't realize at the time, and has become more essential in the multi-GPU setups is the per-rail output ratings. The only option at the time was to go "bigger", but that just upped the rating for each rail.

So I checked the online ratings from Antec for this power supply. You can see in the "Output" chart that the 12V line had a max rating of 28A, easily handling the requirements. The sticker on my particular power supply notes that the max output is 20A, startlingly close to the minimum requirement for this power supply. Why bring all this up? the machine would do a hard lockup when I hit certain intense gaming situations. Knowing all this, you can understand why I decided to buy a new power supply.

Getting back to power supplies, the vendors have adapted to the higher requirements in multi-GPU systems by adding additional, independent 12V+ rails. This way the graphics card can be given its own rail with a 20A (or higher) rating. A quick check to trusty ncix.com showed that there are several suitable power supplies with multiple 12V rails available for ~ $100. The Antec NeoHE 500 power supply is a dual-standard EPS12V and ATX 2.0.1 power supply with 3 independent 12 V rails and an efficiency >= 85%.

As an interim solution, I've under-clocked the video card (using the CoolBits registry hack to unlock the NVidia driver). I believe that should keep it away from the brink. Probably this was what allowed my older 6800 GT to work reliably. But now I need to get something to ensure real reliability.

Computational update

I'm not doing too spectacularly in the "blog every day" department, but I thought I'd write up something about my computer. I haven't done anything about my video card, even though I have an RMA number for it. Maybe I'll call them back today and point out that I haven't gotten a mailing label, just the same RMA form as before. This form says that in the US, use the BFG Tech corporate UPS account to get things sent back, (you just have to supply the packed box). This does not apply to Canada, so if I don't get a pre-paid mailer or promise of financial compensation, I'm not sure what is up.

Other than that, things seem pretty good with the ol' machine. I've had some Win32 services die on startup, but no blue-screens. The same old punishment continues, so I guess things are okay.

In the punishment department, I've loaded up the "Insurgency" Mod for Half-life2. It's basically Battlefield 2 on the Source engine. So far, pretty good, despite the beta status. I'll be much happier when I get the better video card as this mod is very "realistic". That means no maps and no coloured tags over the players. You have to recognize that the little blur down there is a friend or foe on your own. The worst bugs I've found are in the squad system - there is no canceling and there is no easy way to switch types if the squad is full. However I've just randomly hit a number and been given the class even though the slot is full. There are too many specialist slots (rpg, sapper on insurgents) and not enough simple rifleman.

Anyway, that's the update for now. Still haven't brought back my old RAM for warranty stuff, but there is still time.

Moderate success!

I have complained, to all who would listen, about my computer troubles over the last few months. It has been most annoying and interfered with my gaming, er blogging. So I thought I'd take some time to describe what has been going on.

A few years back, I invested in a new system. Althon 3500+, MSI K8N Neo2 Platinum (Nforce3 250 Ultra chipset), OCZ PC3500EL (2x512 MB match pair), 250 GB Seagate HD (7200.8 SATA interface), DVD+/-R CD-RW/R, Antec Sonata case, Antec TruePower 480 PS, BFG 6800GT OC. At the time, pretty cutting edge without breaking the bank. In recent months, instability has crept in. At first, I thought it was something with the North Bridge - specifically the fan on it. The AGP slot goes right over the fan and 6800 GT is known for its prodigious heat production. This suspicion was later confirmed when the fan started making noise. Being lazy, I just let it be with blue-screens and lock-ups happening frequently.

Then the video card failed. Streaks of random colour appeared across the display. I must take action - no one can use the computer like this! I called BFG, who have a life-time warranty on their products. They determined, over the phone, that the RAM on the video card had an issue and I would have to send it back. They gave me an RMA number and I sent it back. They have excellent tech support, when you can get through, btw. Concurrently, I ordered a new heatsink/fan combo for my north bridge and some fresh thermal grease. Then I pulled my MSI Ti4200GT from my old machine to continue hobbling along. One can't be entirely without the Inter-weeb, don cha ya know.

The new parts and the replacement card arrived a week or so later. I order a Vantec VGA/chipset kit. The heatsink/fan combo was intended for older video cards to replace a passive heatsink with a copper-based one and integrated fan. My motherboard used the same fan/heatsink layout so this would be a drop-in replacement. Unfortuneately, I had to remove my motherboard to pop the retention pins out from underneath. The first thing I noticed was that the original heatsink/fan was so light - barely any metal at all! That and the reason the fan wasn't working was because the intake fins were clogged with dust. Ugh. The replacement heatsink/fan was copper and had some heft to it so I just made a mental note to clean that out every month or so. Replaced the motherboard, swapped in my cooling tower heatsink (a Thermaltake Silent Tower) and the replacement 6800GT. Everything fired up nicely, much much cooler than before, but the machine locked up hard when XP was started. Eventually tracked it down to the replacement card - it worked fine in VGA mode (during boot, in Windows XP Safe Mode) but as soon as the drivers tried to initialize it, the entire machine locked up hard. I swapped in the old Ti4200GT and away we went. The random lockups are done!

But they weren't or course. The problems that I attributed to heat weren't really the cause of all the problems. I finally fired up UBCD (Universal Boot CD) with memtest86+ and quickly discovered that one of the OCZ modules was producing constant errors. I swapped the locations to see if it was a motherboard issue, but the problem moved to the other bank. So off I went to get some replacement memory. I was tempted to get more RAM - move to 2 GB from 1 GB, but money reality really was suggesting that I should do a simple replacement. The best local price I saw was about $39 for a single 512MB stick, and I intended to buy two to fully replace what I had. I ended up going to KoolComputers (Wharncliffe store) and was surprised when the store clerk dug up some OCZ modules! I said that I had a matched pair of OCZ that developed a prolem and she indicated that they deal with OCZ all the time. If I wanted to use the lifetime warranty on the OCZ modules they would be able to do it on my behalf. Sounded good, but I needed something now. The replacement modules were slightly lower spec (PC3200 vs PC3500) but they had the nice heatspreaders and were $49 each! Awesome. Went home and ran them through memtestx86+ again, with no errors. So it has been half a day and so far the reliability has been high. I'm hoping that's the source of the reliability issues. Next up, performance!

That leads right back to the 6800 GT. I had spent 28 minutes on hold with BFG waiting for an operator. Especially infuriating was the point 5 minutes in when the system informed me that it would be about 3 minutes until the next call. After waiting an additional 20, I hung up and called back. That time the voice said 16 minutes. Since I didn't want to wait until next week, I left to go get my replacement RAM. I called back on return and got through immediately. The tech listened to my description of the problem and said that I'd need to send it back. That was okay - I expected that. What I didn't expect was the "Since the card we sent you was defective, we'll send you a label so you don't have to pay to send it back." Awesome again! The call was done in 11 minutes, punctuated with several minutes silence while the tech consulted various things. I look forward to the next round with the video card.

This brings me to my experiences with BFG Tech support. I have been the guy answering the phone before and I've called other tech support lines, although very rarely. The BFG support line has been the best I've experienced, for one main reason - they listen to what you say first and then figure out the next step. When they ask what the problem is, I summarize the problem and describe what I did to fix it. This is usually enough for the tech to determine that I've done all the simple things (reboot, try new drivers, tighten connections etc etc) and more complex things (replace card with a known good card and what happens). This last call and the call where I sent back my original card were excellent examples. They listened and when I was done, the first words were "well, sounds like you have a defective card, we'll set you up with an RMA." I was so pleased that I wasn't told "so let's try using the latest drivers" or some-such.

Anyway, all you people (okay, the one guy who reads my blog on occasion) have been treated to an extra long edition. I will try for the "more often, less long" philosophy in the future, but I've been starved of writing entries of late.