[CP5670]

At the risk of creating yet another of these dreary threads, I finished running stability tests on my new computer configuration last night and tried out several games with it. I took a few pictures while it was running through the tests, since watching Prime and the other stuff go on for hours can get pretty boring:

http://home.comcast.net/~cp5670/c1.jpg

http://home.comcast.net/~cp5670/c2.jpg

http://home.comcast.net/~cp5670/c3.jpg

http://home.comcast.net/~cp5670/c4.jpg

The cables are in something of a mess but I couldn't get things much neater.

http://home.comcast.net/~cp5670/c5.jpg

http://home.comcast.net/~cp5670/c6.jpg

http://home.comcast.net/~cp5670/c7.jpg

http://home.comcast.net/~cp5670/c8.jpg

This should last me for another year, although I'll probably upgrade the memory in a month or two.

Some thoughts on SLI:

In a nutshell, it takes a fair bit of work to set it up and optimize it for each game, but the results are worth it. There are a couple of different SLI modes and you have to try and see which one has the best performance, has no bugs, works with vsync and so on for any given game. The third party nHancer utility is basically a must have for setting up the profiles necessary to use SLI and making full use of the AA modes (all 17 of them), especially as the default nvidia profiles aren't always optimal.

Far Cry and Doom 3 both work well at 2048x1536 with 4x AA / 16x AF. Far Cry in particular looks like a whole new game; the huge outdoor areas and viewing distances benefit immensely from these settings. There seems to be some bug with the nightvision though, which causes big drops in SLI performance and is supposed to be fixed in an upcoming patch. SCCT with its HDR is more demanding, but the framerates are quite good at 1600x1200 with 16x AF, never dropping below 50. FEAR also does pretty well at the same settings, but Call of Duty 2 remains a total performance hog; instead of 640x480 on my old card I can now play at 1280x960 with no AA, despite the 90% gains from SLI over one card. I might try using the DX7 mode, which is supposed to triple the framerates while having a minimal impact on image quality.

I tried some older games as well, where the SLI AA feature is very useful. FSO and Deus Ex both get an excellent 85fps at all times at 2048x1536 with 16x SLIAA and 16x HQAF. Descent 3 works perfectly at 1600x1200 16x SLIAA / 16x HQAF, but the game menus don't support any resolutions above that. Anyone know if it's possible to force a custom resolution in this game?

As for noise levels, the cards are very loud on startup (although it's really an air rushing noise rather than any motor hum) but the drivers quiet them down a lot once Windows has loaded up. They are still a little noisy by default, but it's possible to set the fans to run at their minimum 25% speed in Windows, at which they are inaudible to me, and ramp up to the normal 48% speed in games only. Overall I don't consider the noise a problem with the appropriate tweaking, but people say the normal GT fans are louder.

Overall I'm pretty satisfied with this especially for the price I paid, and am going to play through several of these games a second time with the improved graphical settings.

[Roanoke]

and here I am struggling with 800x600 LOl. What cards are you running BTW ?

[Dranon]

They look like EVGA 7800 GT's... although which iteration, i'm not sure.

They are the cards i want to get though once i build my next computer.

[Stealth]

yeah they're eVGA 7800 GTs.

the only thing better than dual eVGA 7800 GTs in SLI, is dual MSI 7800 GTXs in SLI

[Stealth]

by the way.  looks good CP:  that's a pretty neat looking power supply.  i'm going to try it on the next computer i build.

guess there's a computer out there that can compete with mine now

naaaaaaa

[Dark RevenantX]

Just crank up your speakers if it gets loud.  The game noise always drowns it out.

[Stealth]

or get headphones

[CP5670]

I've gone through almost all of the games I have now, including all the old ones. Almost everything receives a sizeable boost from SLI at a sufficiently high resolution and AA setting. The only exception I've found is Splinter Cell Pandora Tomorrow, which gets no benefit over one card in any SLI mode. It's strange since it uses the same engine as the first Splinter Cell, which works fine.

By the way, as I was going to check the COD2 scores on Firingsquad and make sure that my performance was in line with what's expected, one headline on the front page caught my eye:
http://www.firingsquad.com/news/newsarticle.asp?searchid=9094

I guess this has become big news.

and here I am struggling with 800x600 LOl. What cards are you running BTW ?

lol, I was forced to use that in some newer games on my 6800GT and was getting tired of it. They're EVGA 7800GTs as others said, the CO versions (the whole thing is listed here). There were some big sales on these cards around three weeks ago and for $290 each I think they were a good buy.

the only thing better than dual eVGA 7800 GTs in SLI, is dual MSI 7800 GTXs in SLI

They're actually pretty much identical at stock speeds, as the MSI cards are reference speed models, but the GTXs will generally have more overclocking headroom.

Have you run into any games besides SCPT that won't cooperate with SLI?

by the way.  looks good CP:  that's a pretty neat looking power supply.  i'm going to try it on the next computer i build.

It's an OCZ Modstream 520W. I'm not sure if I would buy it at the current prices, but it was a great deal compared to other models when I got it back in February for $90. It's probably still the best modular one though.

Just crank up your speakers if it gets loud.  The game noise always drowns it out.

I actually use closed-ear headphones for all my gaming, so I don't really care about the fan noise during games. The only issue was when I'm just doing stuff in windows, but I can't hear the cards over the CPU fan at that minimum setting so it's working out.

[Stealth]

They're actually pretty much identical at stock speeds, as the MSI cards are reference speed models, but the GTXs will generally have more overclocking headroom.

naaaa, the GTX will outperform a GT any day.  Again, there's no conspiracy here... there's a reason you pay $200 more for a GTX LOL ... it has less pixel and vertex pipelines and a slightly slower core and memory clock apparently.

I guess it boils down to whether you're willing to pay the extra few hundred $.  nonetheless, dual 7800 GTs will still 0wn , especially for gaming.

[CP5670]

Check this out:
http://www.firingsquad.com/hardware/evga_e-geforce_7800_gt_co_edition_review/

The reference GTX's lead is under 3% in almost all of the benchmarks, 6% in the remaining cases, and although it's not shown here they fall behind by about 3% in Doom 3 and Quake 4 (hardwarehell has scores of these games, but their site is currently down), so they're basically comparable. At most hardware forums people always recommend these cards over two GTXs or a single 512 GTX because of this. The GTXs will pull ahead once overclocked since they have a lot of headroom there, but at reference speeds they're a poor choice for $900, so go and overclock those cards; it's the first thing I did with mine.

One somewhat annoying thing I have noticed is that the vsync support in SLI does not quite work correctly in SLI AA mode despite Nvidia's claims, which can be a problem since some older games can only make use of both cards in this mode and show tons of tearing without vsync. Some games like the Unreal engine ones and Descent 3 have a software-based framerate limiter that can be used to get around this, but FSO doesn't have any such thing to my knowledge. Something to bug the coders about, I guess.

[Stealth]

Check this out:
http://www.firingsquad.com/hardware/evga_e-geforce_7800_gt_co_edition_review/

remember:  don't compare your card to the stock GT... the CO is just a beefed up version of the GT, making it a closer comparison to the GTX, but if you overclock that GTX a bit, it'll be the same difference between it and the CO, as between the stock GTX and GT.  i mean don't get me wrong, the GT's still a great card, especially since you got eVGA's version... but it still can't compete to a 7800 GTX

also, i'll overclock them in a little bit.  i want to get my airflow and air circulation in the case worked out first

[CP5670]

I guess you didn't look at the scores in that review. I'm only considering the EVGA GT version and a reference GTX, which are indeed very close as I said earlier. The GTXs might have been worth looking at if there was any significant (greater than 10%) difference at stock speeds, but for $900 you might as well pay a little more and get the considerably faster X1800XTs. You're right about overclocking though, since maxed out overclocks on an average CO and an average GTX would increase the gap between the cards, although not beyond 10%.

The temperatures actually don't make much of a difference for video card overclocks by the way; most people report that their max overclocks don't change a lot when going from the stock coolers to much better aftermarket coolers like VF700s.

[Stealth]

i did read the review...  the review is concentrating on the eVGA GT CO, a more souped-up version compared to most GTs (the bar highlighted in yellow), but i still see pretttttyyyyy big differences between a 7800 GTX and 7800 GT. 


etc.

seems (as you said) that some games and settings the 7800 GT is almost identical in specs to the 7800 GTX, but i mean, in the above two images, the GTX is dominating the GT.  10%+ EASILY. 

just depends whether you can afford to spend $400 more on dual GTXs compared to just dual GTs

EDIT:  you DO have the eVGA GT CO, right?  which of the three eVGA versions of the 7800 gtx do you have?

[CP5670]

Have you computed the percentages? We're both a little off on the first one, which comes out to 8.3%, but the second one is only at 5.2%. And I would put more stock into the second one, as the first isn't quite a playable setting on any of the cards. If you consider that dominating, you should take a look at the 30-50% jumps across the board with the 512MB GTX.

Anyway, I've been playing some more games today and the vsync thing is starting to increasingly annoy me in actual gameplay, and it's not just limited to SLI AA mode. In many cases I can get a game working great without vsync but has performance issues with it on, including several newer ones; I am noticing the tearing a lot more when playing normally than when I was just trying out stuff earlier. I'm surprised this is so rarely brought up as an issue, and I have the refresh rates at 85hz or 110hz; on an LCD with 60hz, which is what most people use, the tearing would be intolerable.

Are you getting irritated by this too? This is actually bothering me much more than it ever did on my old 6800GT, since the framerates are exceeding the refresh rate all the time. I might try the leaked 82.04 drivers and see if they do anything over the 81.94 set.

The good thing is that I got these cards at such a good price that I could sell them on ebay at a profit. I think I might keep the cards for a month, then sell one card and use EVGA's step-up program to trade the other card in for a G71, if that comes out when it's rumored to.

[Dranon]

must get tax return....


What monitors are you guys using to get that high of a res?

i'm only one a pansy 17"... but will most likely upgrade when i build my next comp in a month or two...

[Stealth]

Have you computed the percentages? We're both a little off on the first one, which comes out to 8.3%, but the second one is only at 5.2%. And I would put more stock into the second one, as the first isn't quite a playable setting on any of the cards. If you consider that dominating, you should take a look at the 30-50% jumps across the board with the 512MB GTX.

are we looking at the same specs?  i'm comparing the GTX to the GT, NOT to the GT OC, which is just an overclocked GT.  The differences between the GTX and the GT are a good 15%.  if you want to compare the GT OC (yellow bar) to something, compare it to an overclocked GTX, which isn't shown on the graph...

i think it's amazing though, if you look at the 6800 Ultra.  6 months ago the Ultra was still a top of the line card, and i remember a year ago the coolest card you could buy was the famed and fabled "6800 Ultra", which took up a whopping TWO PCI slots!!!.  now look how the 7800 gt leaves the ultra in the dust

Also, i never play with Vsync except in CSS, for some reason.  let me start playing with it, and get back to you.  i'll bet we'll have the exact same problem.

[CP5670]

I'm comparing the CO GT and the reference speed GTX; not sure why the normal GT is relevant. The CO is an overclocked GT, but that doesn't really matter since it's set to that speed by the manufacturer themselves using cherry-picked cores and memory, so 470/1100 is its stock, out-of-the-box speed. You can overclock it further like it any other card. I have both of mine running at 500/1150 now, although they seem to run okay up to about 530/1200. You can also get GTXs with similar manufacturer overclocks, but you will have to pay more than the $450 price of reference cards for them.

But yeah, video cards are moving forward at an awesome rate. The rest of the PC industry has stagnated a lot, with no major advances in CPUs, memory or hard drives for at least two years, but video cards have become over five times as fast in that period (9800 Pro to 512MB 7800GTX), and that's only considering single card setups. The R580 and G71 look like they're set to continue this trend at the end of next month.

Let me know what your experience is like. Since I'm thinking of returning or selling one of these cards and just using the other one by itself until the G71 comes out. For the $580 I paid I expected better than this, especially as Nvidia themselves advertise that vsync works in all games as of the 81.85 drivers.

What monitors are you guys using to get that high of a res?

i'm only one a pansy 17"... but will most likely upgrade when i build my next comp in a month or two...

I'm using a Mitsubishi 2070SB. It's arguably the best consumer monitor out there (aside from the professional ones that cost over $5000), as it's the only one that both does 2048x1536 at 85hz and has an extended brightness range tube. Unfortunately they were discontinued last year and are nearly impossible to find now, new or used. I'll probably stick with mine until SED monitors come out in 2007, which are looking like they will kill everything else out there.

You can still get other high end CRTs at great prices on ebay (the Sony FW900, another great monitor, is common on there), although there is always a significant element of risk in buying a used or refurbed monitor.

[Stealth]

I'm comparing the CO GT and the reference speed GTX; not sure why the normal GT is relevant. The CO is an overclocked GT, but that doesn't really matter since it's set to that speed by the manufacturer themselves using cherry-picked cores and memory, so 470/1100 is its stock, out-of-the-box speed.

because you're comparing a souped up card, to a stock card.  it's one thing to compare the 7800 gtx to the 7800 GT, but to compare the GTX to a souped up, overclocked GT?  why?  that's NOT a stock GT... that's overclocked, so big deal if it's closer to the GTX's performance...... an overclocked GTX would rape it

Also, by the way, my processor went from 69 degrees after heavy gaming, to 45 degrees.  pretty big jump i think.  and the only thing i changed was the 'stock' thermal compound that came with the zalman, to a "silver" compound (quite a healthy amount too), and reseating the processor.  AMAZING jump in temperature though

[CP5670]

because you're comparing a souped up card, to a stock card. it's one thing to compare the 7800 gtx to the 7800 GT, but to compare the GTX to a souped up, overclocked GT? why? that's NOT a stock GT... that's overclocked, so big deal if it's closer to the GTX's performance...... an overclocked GTX would rape it

uh, yes it is a stock GT, just not a reference speed GT. There is a difference. All cards are stock cards by definition unless overclocked by the user. The thing is that this card costs about the same as the reference GTs if you look around a bit. I'm not sure what you're trying to argue here.

Also, by the way, my processor went from 69 degrees after heavy gaming, to 45 degrees. pretty big jump i think. and the only thing i changed was the 'stock' thermal compound that came with the zalman, to a "silver" compound (quite a healthy amount too), and reseating the processor. AMAZING jump in temperature though

The arctic silver 5 stuff? I doubt the compound alone made that kind of difference; it probably had much better contact the second time around. Even the difference between the stock AMD thermal pads and AS5 is 4-5 degrees in most cases. Good to hear it though.

[Descenterace]

Those thermal pads are rather thick, though. If applied correctly AS5 should only fill the gaps; most of the heatsink should be in direct contact with the heatspreader.

Lapping a CPU that's running very hot and replacing the stock pad with AS5 can drop the temps by as much as 20C in some cases, as I know from experience.