while i was still going to school at UCSD, they were building Cal(IT)^2, the California Institute of Telecommunication and Information Technology. now that it's done, i've toured it (on work time... i felt it was important to see the campus's resources and facilities hehe).

it has this awesome thing called the HIPer Wall... i felt like i was on a holodeck



i was wondering if that's what you meant. and yea, i'd tend to agree--it'll happen (probably sooner than later).



my thoughts exactly. i usually game with winamp running in the background, and it is pretty disappointing when i have a sudden urge to listen to a particular song, but don't want to alt+tab into the other screen. i've also been playing morrowind lately (it was a gift, and since i was thinking about getting oblivion anyway, it seems like a good 'preview'). i'm kind of annoyed with it because i feel sort of lost (i'm used to the final fantasy line of RPGs with their tightly knit plots). i'd love to have a second monitor i could use to surf walk-thrus and such.

in fact, the only reason i think i might want a "wing" monitor on both sides of my primary is my neurotic need for symmetry: one would functionally be enough i think, but IMO a third monitor wouldn't be totally extravagent either. i saw several decent-looking monitors for right around $200 when i did a quick browse on newegg for my last post (of course, if i want to use something other than the stock monitor stand, it'll cost me).

also if i ever did want to try spanning multiple screens for a video game, i'd much rather do it with an odd number of monitors. not a big deal for productivity, but i hate the idea of a seam running down the middle of my viewing area in a game.

Back when I was making my choices for this computer, I chose an acer X222W 22 inch Widscreen DVI flatpanel.

widescreen really is nice. it doesn't seem like a big deal until you have one and get used to it.

ugh, i'm finally getting around to actually cleaning my room since building my computer. i've had little bags of screws, extra modular cables, OEM stickers, proofs of purchase, and all that crap laying around. funny how knowing my mom will see my room later is a big motivating factor.

i finally started tweaking some of the advanced BIOS settings. i wanted to get my RAM timing to 4-4-4-12, since that's what i paid for (my BIOS defaulted it to 5-5-5-18).

when i enabled the advanced BIOS features, a field appeared called "refresh to ACT delay." the stock value it had was 42 (IIRC), but in the advanced column it had a zero. before changing anything, i googled the phrase, and found a forum where someone suggested setting it to 100. i tried that, and it seems to have worked fine.

but what does this actually mean? should i try lowering it back to the default value? should i try getting it even lower? unfortunately there wasn't an 'auto' option.

anyway, any help would be appriciated. this overclocking stuff is interesting, but way more complicated than assembling hardware components.

---

as a totally separate item, Craig, i was wondering why you were down on nVidia chipsets? i mean, i'm not really interested in SLi, but they don't seem like bad chipsets. is it stability issues?

My motherboard is the ECS NFORCE4M-A v3.0, socket AM2, with a maximum RAM capacity of 32 GB DDR2 800. It is a workstation board. FDD, 4 PATA devices, and 4 SATA150 hdds. I doubt that I will ever see 8 GB DDR2 modules, although I can dream and drool.

i know exactly what you mean

i'm really happy with what i have, but you wouldn't know it from the way i'm constantly browsing computer parts (and reading up on computer technology, and reading about overclocking). i'm a very curious person by nature, and any time i can create or assemble something, i'm in heaven.

my daydream du jour is that i'll meet some rich person who wants me to build a system where money is absolutely no object

It's not that I'm down them because I'm not really. I've used several nForce motherboards in the past. My only dissapointment is that I didn't see a big gain in memory bandwidth going from a DDR400 to DDR800 system (according to my own benchmarking). I was expecting to, but didn't. OTOH, the P31 provided those big gains I was expecting to see. So, for a chipset that advertises top notch performance, it didn't deliver.

Regarding memory timings, I think it's a good idea to use a benchmarking program to see if changes are actually worthwhile. I've been using Sandra and PerformanceTest. I also use MemTest to check memory is stable after making changes.

I think you'll find that tightening the timings doesn't provide that much of a performance increase. I've found that increasing the memory bus speed improves memory performance more than anything else (by running up the host clock which also increases FSB and CPU core frequency). In that case, the more relaxed timings in auto mode are better for stability. Check it out and see for yourself.

I also had a question about that refresh to ACT delay setting. It should be at some value and I don't know why the BIOS defaults to zero when in manual. I set it to the value indicated in the default column. Googling it found people setting it to a range of values, but I didn't spend a lot of time on it. I'm leaving the timings on auto and only running up the host clock. I could probably squeak out another 5% in memory bandwidth by tigtening the timings, but the increased host clock gives me much more than that. I'm happy with what I'm getting and running the timings in auto is less trouble.

When I built this system, I had to settle for two 2GB kits of Kingston ValueRAM, for 4GB. This is not enough when working with Vue 6 Infinite 64 bit, and working with ecosystems. The ecosystem editor slows to a crawl. I just replaced one of those kits with a G.Skill 4GB kit (same timings as the Kingston), and will replace the other in January for a total of 8GB.

What I wanted to build, was a dual, dual core Opteron system. Unfortunately the main parts cost, not including shipping was $1,986.00 for mobo, 2 opterons (275 or 285), 8 x 1GB DDR400 ECC Registered, video card, and 700+ watt EPS PSU.

So, I spent $550 on software and $1,450 on hardware, including shipping.

A Q6600 setup would have cost less than the opteron one, but, was still out of my budget at the time.

Just a heads up: ram prices are going to increase dramatically in the near future as manufacturers cut production to bring the price up. I've seen $75/gb forecast for 2008, so upgrade now if you can because ram is as cheap as it's going to get.

I just picked up 2gb of OCZ Platinum DDR2-800 for $17 after rebate Check out a boxing day sale near you.

i see - yeah, i'd have to say real performance trumps advertised specs any day of the week.

there's also this voice in the back of my mind that suggests you'll get better performance by matching your chipset and CPU (AMD with AMD, Intel with Intel, which kind of leaves nVidia out in the rain). it's definately founded on no more than a personal, "common sense" assumption, and it's probably a bit superstitious.



thanks for the recommendations are you using the Lite version of Sandra? i visited their site, but got sidetracked before downloading anything.

so far i've downloaded Prime95, Intel Thermal Anaysis Tool, CPU-z and set up a PC Pitstop account (i'd provide links, but google should be all you need).

CPU-z seems a bit wonkey - it refuses to read my CPU multiplier above x6, even though i've made sure it's at x8 (per stock settings), and as a result it shows my CPU speed 25% lower than it should be. i triple-checked in BIOS, and when i run Intel TAT, it show my CPU speed correctly at 2.6 GHz.



fair enough - right now i've got my memory multiplier set to 5:6 so i get the full 400MHz on RAM w/o OC'ing the CPU - out of caution. i wanted to familiarize myself with things a bit before trying to boost my performance. i've never OC'ed before because i've always had a Dell/HP/Gateway (does Gateway still exist?).



okay, i did a little more research on wikipedia and found a very, very small blurb:

"The auto refresh command also requires that all banks be idle, and takes a refresh cycle time tRFC to return the chip to the idle state. (This time is usually equal to tRCD+tRP.)"

tRFC = refresh to ACT delay, at least according to a blurb i found on a random forum... which could be incorrect. assuming it is in fact correct, you'd want to add the middle two values in your primary RAM timing to determine this setting -- if you leave your timing at 5-5-5-18, the usual setting here would be 10. hey, i figured something out all on my own!  well, sort of. i still don't know what tweaking it would do to a system; it's really only superficial knowledge.

edit: after reading a bit more, it seems like few users can actually get that value down that low... after further reading, i'm realizing that tRFC is the next field (and the auto settings put that at 10 anyway).



i'm in no way involved in 3D modelling, but i think it'd be a lot of fun for me if it weren't so expensive. i'd never be able to justify spending that much money on advanced software and hardware just for "a lot of fun".



hmm, is that all RAM, or DDR2 specifically? i remember reading that DDR3 doesn't present any major gain to users, but it's advantageous to manufacturers (something about it being easier to design and produce). maybe this is the beginning of a supply-driven shift towards DDR3?



Canada's only 10 hours away by car 

i've thought about throwing in an extra 2x512MB in RAM, but it actually seems to have been more expensive than 2x1GB. i've got XP, and i know it won't recognize more than a total of 4GB system memory, including graphical RAM and swap file. my ideal in XP would be to have no swap file at all (say, 1GB of graphical memory + 3GB of RAM), but it doesn't seem economically worthwhile.

I haven't found that to be the case really. Although I have to admit, there is a certain comfort factor involved in running a chipset and CPU from the same maker. I've been happy with various chipsets from nVidia and VIA over the years. The nForce 520 has been the only exception and it may have been due to a quality issue with the ASUS board it was mounted on.

As far as AMD goes, they've never really been big into making chipsets the way Intel has. Pickings have always been pretty slim for motherboards based on the relatively small number of AMD chipsets that have been released. That may change for the Spider since it's marketed as a complete AMD platform. Unlike Intel, AMD is a major performance GPU maker with their ownership of ATI. Intel does offer some GPU products, but they don't compare with the high-end stuff from ATI and nVidia.


Yes, the lite version is free to use. It doesn't have all the utilites of the non-free version, but it has the stuff you need to benchmark your system.


Instead of Prime95, try this version of Orthos Prime. It's a nice front end for Prime95's torture test and initiates two threads by default. Unlike Prime95, it allows you to fully load both cores.


The Gigabyte board makes it really easy. If you run up a setting two high, the BIOS will detect any failure and reboot with stock speeds. Give it a few seconds though. It does a complete power cycle when it happens. That kind of freaked me out the first time. One thing that's handy is the F11 and F12 function in the Gigabyte BIOS. It allows you store and recall your own setting profiles. I'm using it quite a bit.

As far as damage risk, it's the over-voltage you need to be concerned about. You can pretty much run up frequencies without concern. However, with voltage control in auto, the BIOS will increase voltages as frequencies are set higher. You may want to keep voltage control in manual.

One note about memory voltage, the Gigabyte BIOS does not indicate the actual memory voltage. It only shows increments of .1 volts in the voltage setting. I had no idea what it was running at. The only way to get an indication is to install the Gigabyte motherboard utility. It shows the memory voltage there and that's the only way I've found to get it.

Well, I have to wait until the January 3 (Social Security Disability Insurance).

Dystopic,

The upgrade price for me was $299 from Vue 6 Pro Studio. Vue is 3D environment. Modeling is done in other programs. There is Easel, Esprit, Pro Studio (Esprit + all extras modules), Infinite, and XStream (plugin version of Infinite). 32 bit is stuck with a 2 GB RAM space, while 64 bit is not.

Galciv and Galciv 2 run fine in XP64.

3D CG is my primary interest.

okay... i'm sure i'm just naive and ill-informed, but... i read that online somewhere Vue is 3D environment, and i thought that meant you could use it to model, you know, trees and mountains, rivers, maybe cities...  i take it now that's not the case. so what does "environment" mean when you say "3D environment"?



i figured it was probably superstitious, but i'm glad i'm not the only one who perceives some small comfort factor



i think it will. i remember a year, year and a half back, when i learned that AMD had about ATI, i was talking with a friend who said he was really interested to see what they'd do.

as we've both said already, the Spider platform in its current form only looks promising. i wouldn't actually want to build one now though. the 790FX looks like a really solid NB, but AMD could stand to produce a better SB to go with it. their current boards with the 790FX still use the SB600; they're expected to release the SB700 at about the same time the HD 3870FX is released, Q1 next year. hopefully that will shore up the Spider platform to its full maturity; i really look forward to seeing these setups in action.

Yea, it takes some serious horsepower to do that stuff. I've read the fastest machines around right now can render about 1/4 the speed required to do movie quality stuff in real time. Applying Moore's law, it shouldn't be too long until a peformance gaming machine can do that sort of thing live. One of the things that motivates me on these upgrades is that I enjoy experiencing the improvement in 3D video quality as machines get faster. I imagine a game with video quality that compares to the Toy Story movie.

have you ever heard of machinima? i stumbled across this concept/phenomenon reading an article about the growing inter-relationship between academia and the gaming industry. basically, it's when people take a 3D engine (usually from a video game) to render cinematic stories in real time.

with the way game engines are developing, this could give some really powerful tools to would-be home CG storytellers and filmakers. sort of like next-gen cable access TV... for better and for worse.

You know, I was thinking the other day, why don't they package GPU's that same way as CPU's and put a socket on the motherboard for it. Wouldn't that solve a lot of problems? It sure would be nice to use a good CPU HSF instead of that hot noisy card mounted thing.

Thermalright to the rescue again. The HR-03 just won HardOCP's video card heatsink roundup. It was effectively silent with a big lowspeed fan and even managed to cool a stock clocked 8800GT passively (although just barely).

i had some similar thoughts myself; i want to step back a second to answer the question (why they don't do that).

in social science, there's a loosely defined concept one of my professors called "historical inertia." basically, once enough people with enough power are invested in a certain paradigm, it's hard to change that paradigm. the QWERTY keyboard is the classical example. it was originally designed to be difficult to use because in old-fashioned typewriters, the punch levers could get locked if keys were pressed too quickly. even though that problem was eventually solved, and there are numerous alternative to the QWERTY keyboard, it's still the standard because so many people already know how to type on a QWERTY quickly, and it's easier to stick with an inefficient design you can use efficiently than to teach yourself to type all over again.

it's the same issue with basic component design. lots of alternatives are possible, but none of the problems with the current design are serious enough to warrant a change. just look at the BTX design, which was supposed to solve a number of problems with the ATX design. but it never really took off, and i think the biggest reason is that so many people already had ATX cases, and manufacturing techniques were more oriented towards ATX designs (though if Intel had marketed BTX more aggressively, it could have probably done better).

now if you want to talk about re-designing MBs from the ground up, there are a few things that occur to me. could motherboards be made to be modular? imagine buying your MB in pieces, and connecting those pieces with something akin to a PCIe bus with a right-angle orientation. assembly might involve some extra screws to keep things secure, but i don't think that's a big deal.

i walked down this path of thought because i sort of wish i'd got a board with a southbridge that supports RAID. but i don't want to upgrade my whole mobo just so i can play around with RAID. then i wondered why NBs and SBs don't have a socket mounting system; the obvious answer is that it wouldn't necessarily be an advantage (even if your new NB supports two PCIe x16 slots, your board needs to be wired to two PCIe x16 slots to make use of the improved NB).

you could get around that in a couple ways. for example, there are a number of motherboards that have one or two PCIe x16 slots that only have 4 or 8 lanes actually assigned to them. if you could dynamically assign those lanes in BIOS, it could make the idea of putting the NB on a socket more appealing (instead of soldered directly to the board). so build all 7 expansion slots at x16 size and number them, put the NB on a socket mount, and allow the user to assign actual lanes to the numbered x16 slots in BIOS.

the next logical step in my mind was - well then heck, why not just make the entire motherboard modular? this sort of design would allow so much more customizablity... (wait for it)... across the board (sorry, i couldn't resist the pun). but seriously, you could more highly customize not only hardware specs, but also form factor, and the ability to upgrade more gradually. i'm sure the greatest disadvantage would be overall cost and the complexity of building a system; single, integrated boards would be the preference of OEM manufacturers. but for enthusiasts and professionals, i think it could present major advantages. those groups generally are knowledgable enough to handle the extra complexity, and for them it might actually save money or (hopefully) at least reduce the amount of money you'd have to spend all at once.

so... waddaya think? crackpot idea, or is there some merit to it?

And those are *very* major limiting factors. I'd hate to see the compatibility nightmares that would arise in such a system. Hardware companies seem to have a hard enough time making decent drivers as it is...


It's probably better that graphics remain behind a more generic interface such as PCIE, particularly for upgradeability and interchangeability. If they were on the motherboard via a socket, chipmakers would probably be a lot more likely to use varying proprietary socket types, with all the fun that involves (nonportability, forced upgrades, etc). Though AMD may go in that general direction with their Fusion concept, I'm not sure if that'll be a real improvement in so many ways.

fair enough; i completely glossed over how complex and difficult creating BIOS for such a system would be  not to mention that you'd probably have to flash new BIOS for every single upgrade (and while i've never had any difficulty updating BIOS, the various warning i've seen communicate clearly enough that there's significantly more risk than your average driver update).

but what about the sub-idea i'd had, to whit:



that doesn't seem too difficult or complex by comparison. then again, i'm not a computer engineer :NOTSURE: 

Good point, but think about it with respect to CPU's. We have Intel, AMD, and whoever else is still around. Each has it's own proprietary socketing system and it doesn't seem to be a problem. However, adding a proprietary GPU socket would convolute things a second order. There would have to be some kind of socket standard. I don't see how something like that would be so different from standardizing on PCIe or AGP or any other video interface standard we've seen.


I actually still have an older model of that heatsink. I used it a few years ago on a 6800GT video card, passively cooled. It was a real PITA to install and I've since resigned myself to using the stock cooler. There's also some major space constraints with that system. It takes up quite a bit of room in the case and there's a good chance for obstruction with the chipset or CPU cooler. I seriously had about a hair clearance with the CPU cooler. Regardless, it worked beautifully and the reduction in noise was wonderful.

i've been fairly impressed with thermalright's heatsinks, based on what i can see; i really like the fact that many of their models come w/o fans, allowing you to pick and configure them yourself (or avoid paying extra if you want to opt for passive cooling). CoolerMaster has managed to impress me for similar reasons.

notwithstanding those facts, i think next month i'm going to buy this Rosewill RCX-Z4. i'm attracted to it because the fin array is offset to be lower than the center of the CPU socket; i'm hoping it'll line up fairly well with my rear exhaust fan, creating positive and negative pressure over the heatsink component. i also like that it uses a standard 120mm fan, which will allow me to replace it easily in the future if i decide i want to (or if the fan simply dies). at $40, it's also pretty reasonable for my budget (especially since some of the more monsterous HSFs run into the $70 range).

i haven't given VGA cooling quite as much thought; for starters, i'm going to upgrade in a few months (i have a friend at work who's said he'll buy my current 8600GTS 512MB whenever i'm ready to be rid of it). i'm not terribly excited by any aftermarket VGA cooler i've seen. some of them look quite nice, to be sure, but it seems like the most effective ones will either cause clearance issues or cover several other expansion slots.

the layout of the Antec 900 is a little unusual. the PSU mounts at the bottom of the case, and there's a 200mm fan at the top. this makes cooling the CPU relatively easy and efficient. you can also mount a 120mm fan in the center of the case window, which places it almost dead center above the 2nd expansion slot (where single x16 slots are most likely to be). basically, i'm not sure how to optimize airflow across the lower half of my motherboard and VGA.

3D Environment starts with terrains and sky, to which you add vegetation to the terrains and clouds to the sky. You can also add objects like buildings and vehicles, as well as figures (humans and other creatures). Since this is 3D, all of these things cast shadows on other things, like the ground.

In Vue you have Solid Growth Vegetation, which is animatible (wind effects). You also have static plants (objects) which are not. Clouds also animate. You can also set up animations in Poser 6 or 7 and import them into Vue.

Rendering uses the CPU(s), rather than the GPU because everything is described in text and images (materials aka shaders), plus add in raytracing for reflections, Radiosity, and other effects.

The user interface is most often OpenGL dependent, and this is where nVidia reigns supreme.

Those of us who do 3D CG as a hobby learn to go for better hardware to reduce render times, even to the point of setting up render farms.

A poor heatsink that lines up with an exhaust fan is still going to give poor cooling compared to a good heatsink that doesn't line up