It counts for something with benchmarks. The point is, benchmarks have jack to do with real world gaming. If, in any real world situation, a factory OC card does not allow you to increase detail settings or resolution while maintaining a playable framerate, what the hell does it matter if you get a few hundred extra 3dmarks? What return have you got for your extra money? Absolutely nothing that I can see.

An interesting data point regarding PSUs . . . a test of the new Intel "Skulltrail" platform with a pair of QX9775 quadcores, 2 8800GTXs in SLI, 4 gb of ram, 2 big Zalman coolers and 2 hard drives, drew 589 watts from the wall at 100% system load.


And yet, if you believe the manufacturers' recommendations, you need a 750w psu for a system that's basically half of that. And some people are spending incredible amounts of money of 1000 or 1200w PSUs for single-processor systems that aren't even SLI half the time.

Dumb.

3d modelling, animation, and video editing apps will use as many cores as you have, often limited not by the app itself but by the license -- although for the workstation this is often done on a per-socket basis, since just considering the number of cpu cores, a dual quadcore machine would require a render farm license in some cases otherwise.

I wouldn't run Skulltrail on a 650W either, but 750 would be more than sufficient I think. I also wouldn't run a psu at 100% load, but also consider than a lot of them don't reach their rated efficiency until they hit 80%. So having a 750W psu in a system that's only pulling 400W isn't just wasting money on the purchase, it's wasting power as well.

You can buy a little meter that plugs into the wall socket, you plug your system into it to see the power draw. Your utility company might have them, otherwise you can probably get one at Radio Shack or your local equivalent.

i've been thinking about SLi and Crossfire. not thinking about doing it, just thinking about the technologies. they're not a bad technologies, they're just not cost effective (especially if you're just gaming). but still, it's there for professionals and for enthusiasts.

but back to the point, wondering about SLi. three 9800's in SLi on a 780i chipset will the the thing to have for nvidiophiles with money to burn. but among those who do have money for SLi, i've noticed cooling can be an issue. three double-wide cards basically only leaves you with one unobstructed expansion slot (typically a PCIe x1 at the top). these configuration also pose, uh, challenges to cooling. it'll be even worse with the 9800s; from what i understand, they're expected to be two GPUs are two separate PCBs smashed together into a double-wide design. as i've said before, i don't consider this nearly as elegant as ATI's solution, two dies on the same PCB, and i agree with Craig that neither is as elegant as the approach being taken with multi-core CPUs.

in either regard, this though occurred to me. if you're going to do triple SLi with 6 GPU cores, wouldn't it be tons easier to cool them all on a single PCB? i'm imaging this thing being parallel to the motherboard, with its three PCI connectors on the underside. i'm sure economics are the reason something like this would never see the light of day. the market for the highest-end graphics card is slim enough; the market for those who'd want to buy 3 of them probably wouldn't justify an integrated triple SLi solution - or even double (perhaps an integrated double SLi could be as simple as putting the GPU, RAM and other FETs on the top/back of the second, lower card and integrating the cooling of both). too bad for the enthusiasts, i guess.

PS: i got to thinking about this because i was browsing VGA cooling... has anyone else seen the Zalman VF1000 and RHS88? doesn't seem like something i couldn't achieve with individual heatsinks and some thermal glue.

also i read an interesting tip in some random forum. if you're looking for some small heatsinks for some random FETs or something, see if you can find a heatsink on an old piece of hardware. they're supposedly not too hard to cut up with a decent saw or dremel. i've got the low profile heatsink on my old Radeon X600 i might try using for this purpose; my MXZ saw should cut through it without issue.

Man, you had me going there for a minute. Thought I spent a lot more for no reason. But I took a closer look and the model I bought is GV-NX88T512HP. The cheaper model you mentioned is GV-NX88T256HV. The obvious difference is 512MB vs. 256MB of video memory. Gigabyte doesn't list much detail on their products page so I don't know what other differences there may be. It probably has a lower core frequency as well. The model I bought has been deactivated on NewEgg, but I've seen them deactivate products then re-activate them later. It may just be they're unable to re-stock that particular model right now.

i was browsing Gigabytes webpage earlier this morning when i stumbled across and interesting product idea, Gigabyte's i-RAM storage device. here's a link to a review.

basically, it plugs into a PCI slot for power. you plug 4 184-pin RAM modules in, and it presents the memory as a SATA drive. it's got a lithium battery that'll keept he data safe while the computer if off for up to 15 hours. the RAM slots are oriented diagonally, so it'll fit in a standard width PCI slot, though you have to use plain-jane RAM w/o any heatspreaders. it shows improved performance over standard HDDs in the various benchmarks and tests they put it through on hothardware.com. i'd set you back about $130 + $180 for the cost of the memory modules (if you don't have any 1GB modules w/o heatspreaders laying around), but comapred to $700 for SATA SSD's, it's at least worth noting in that respect. it's also worth mentioning that read/write time for SSD storage devices benefit from RAID striping more much significantly than mechanical HDDs, but still... i have to agree with what my friend said. i told him about this thing, and he said it sounds like gigabyte's way of moving old, spare components

if i had the money to eek out slight and occasional performance gains, i think i'd be a little annoyed by what i'd consider a bad design decisions. if i really wanted to create a product designed to replace a standard HDD with RAM sticks, i'd go about it differently. first, i'd design it to fit in a 5" drive bay and (ideally) draw power from a standard SATA connection. the 5" bay should leave just enough clearence to align the RAM sticks perpendicularly (straight up). this could give you room for 6, maybe even 8 sicks on the same component. second, i'd use DDR2, 240-pin RAM. the largest capacity available for 184-pin RAM is 1GB/stick, but for about the same price, you can get 2GB/stick of DDR2, possibly even at 800 speed. i'd also opt for a SATA II connection (especially w/ faster RAM and more of it).

of course, i think it's a little moot with SSD HDD's on their way in. 2.5" SATA SSD's for laptops run in the $700 range for very small drive sizes now, but in the next few years i think we can expect to see them be much more affordable and considerably larger capacities. someone's also bound to come up with one in the 3.5" form factor, which will increase capacity simply by virtue of making more room. the i-RAM is an interesting product in my book, but mostly because i'm guessing it sort of represents an awkward growth phase, or a missing link if you will.

My husband and I would agree with you about Ubisoft.
He has IL2 1946 and it will not run on his Vista machine.
Every ground target he attacks, using my XP machine, he pretends is Ubisoft HQ.

I think solid state storage is a better technology. The only caveat right now is that hard drives are still much cheaper. I'm using 80GB drives that cost $40 a piece. That's cheap! I added an external SATA connector to my computer and use hard drives for backup media now. It's affordable and most convenient. If solid state drives can come close to competing with that kind of pricing, I would have no problem going that route. As far as external devices like those USB memory sticks, I don't have any need for something like that right now, but I can see how it would be highly convenient for sneaker net.

Yea, hard drives used to be one of the most expensive components in your computer. Now they're about the least expensive. It's just wonderful. I bought 500GB drive for DVR and I think I paid $110 for it. Amazing.

I took a look at that article and yea, hehe, things have moved along pretty quick since then. When I played the original Doom in the early 90's, there was no such thing as hardware accelerated 3D graphics. Consoles did little more than Donkey Kong. I remember a few of us gathered around a 14" CRT watching the game for the first time. I thought it was amazing.

I think Crysis refelects the state of the art right now. Things are starting to look pretty real in that game. Though you can still see the polygon effect a little bit. Based on how fast things have moved along, I would bet we'll see movie quality computer graphics in games soon. I think I read somewhere that a current high end system can render about a third the speed required.

Not even close as far as I know. Depending on a the scene and resolution, a high end system would still be many minutes per frame for LOTR level CGI. That's why vfx companies have massive render farms.

Oh well, I guess that's still a long way off then. Let's see, a couple minutes has to drop to 1/30 of a second. So, a system would need to be 3600 times faster. Apply Moore's law to that!


Yea, hehe, some of those Sci-Fi stinkers have pretty cheesy CG effects.