video cards

I am having a new computer made soon, and I am considering my options for video cards in the immediate future, and in the near future after that.

What I'm wondering is, does running CoX benefit, or use the benefit, of multi-GPU video cards? I've heard that some games, if you had something like a GeForce 295, (I think that one has two chips in it?) some games only use one of them. Something about SLI? I'm not sure how to use that acronym.

Anyway, just wanting to make a good decision for my next vid card, and when Nvidia comes out with DX11 cards, I want to be able to pick a good vid card (or two?) at that time.

Thanks for any info y'all can share about how this game uses hardware like vid cards.

if you are buying right now, buy RadeonHD 5x00 series.

If you are waiting till next march, that's about when Fermi based cards will be entering the market... if everything goes right for Nvidia. As it stands right now though, the general hard-core gamer guess is that Fermi's architecture actually isn't going to be that great for gaming. Nvidia's own documentation on the chip avoids talking about gaming applications and talks more about computational (OpenCL, DirectCompute, GPGPU) uses.

If Nvidia's actually able to make a March launch, which some Original Design Manufacturer's and Original Equipment Manufacturer's are really doubtful of, they'll only be launching a month or two before the planned product refresh of the RadeonHD 5x00 series. The high-end video situation has some Nvidia investors wondering if the company will survive on it's own, or if Nvidia is attempting to re-position itself to be bought up as a technology development company. As is, Nvidia has already publically pulled out of some of it's former strongholds, such as chipset development, and is focusing more on it's integrated technologies such as it's ION and Tegra platforms.

The other big problem is that if Nvidia actually does make a March launch with Fermi based cards, they'll be competing with Intel's Project Larrabee. Now, we don't actually know much about the performance and price points Larrabee cards will be launching at. We do know from Epic Software, the guys behind Unreal Tournament, that the Intel cards will likely be targeted at the mid-range $100-$200 market segment, and that the silicon is capable of pushing UT3 at 1080p with the full detail levels.

***

As to your other question... no. The current graphics engine in City of Heroes doesn't actually leverage Crossfire or SLI'd graphics cards setup.

However, we do know in an off-hand manner that the upcoming Ultra Mode does leverage multiple GPU's. The demonstration given at HeroCon was based on Crossfired RadeonHD 4870x2's.

That being said, if you read through the Going Rogue Information thread, it was reported that according to Ghost Falcon, Ultra Mode was capable of running on just a single RadeonHD 4850. The resolution it'd be running at wasn't specified, so it could be that the 4850 could only do it in 640*480. Doubtful... I'm guessing 1680x1050 is more likely given what the 4850 can manage in other games, and Crossfired 4850's can probably hit 1920x1200.

***

As to the final point... the GTX 295 does use 2 cores, and the cheapest one starts at just above $460 after a mail in rebate.

AMD / ATi are lining up a direct competitor, a 5870x2 if you will. Images of it have already been released / leaked. One difference you might notice is that unlike the GTX 295, the 5870x2 is a single slot solution... which means if you really wanted too, you could make a quad-crossfire setup with it for 8-GPU's. You'd be broke... but hey... how cool would that be?

Pricing and availability is still unknown. Because the RadeonHD 5x00 cards are the clear better choice to purchase right now, you'll actually have a hard time buying one. Newegg is completely sold out, although I did find a couple for sale through Pricewatch.

The 5870x2 is likely to be the same story. It's possible that vendors such as Dell's Alienware and HP's VoodooPC will be getting the majority of initial 5870x2 shipments, leaving few for the retail channel.

So, is crossfire/SLI talking about one card with two chips on it? Or is that talking about two video cards, no matter how many chips per card? Maybe I'm not understanding correctly what GPU means, specifically.

If I get a vid card that has two chips, will CoX use them? or will it only use one of the chips?

Does Pentium i7 950 work well with radeon cards?

I've had the impression, over some years, that Radeon cards suffer from teh suk when it comes to drivers. Complaints of messed up ones, and never any updates. Is that the way it is now, with the current high end cards? How do you think it will be for the ones you mentioned coming out soon?

Thanks for the info you shared so far.

Just to add a tiny bit here...

Without going too much into which is better per se, I've never experienced nearly as many problems with my nVidia cards as I have/had with ATI. I'm just posting this to show that everyone is going to have different experiences

On that note, je_saist that was very well put together information. Not just for the OP, but anyone else who might have had these questions. Thanks for putting in the effort!

We'll see....

OMG je_saist, thank you SO much for all of the great info!

Well, I think I'd better get plannin' on using some Radeon cards sometime soon. Wouldn't be fair to never try them after I've found out as much as I have on them.

If you have time for a quick answer, I guess my last question is: are the 5870s worth waiting for (any idea how long a wait we're talking about here?) or should I tell the computer builders to stick in four current high-end Radeon cards?

My current plans are to use a fairly-good vid card in the new compy to tide me over to the next big thing, but if the current high-end crossfire stuff is good enough. . . I guess I'm asking for an opinion.

Thanks, again, for all your shared info, I luv ya, man! (girl?)

Intel doesn't want anyone to make chipsets for the Socket 1366 or 1156 series of CPUs other than Intel. I think this is also true about the laptop versions of those CPUs. This locks nVidia out of the "but my integrated graphics stomps your integrated graphics into last week" laptop and desktop chipset market but they can still sell discrete laptop GPUs.

This annoys nVidia big time and is also now forced to license the magic word to motherboard manufacturers that will allow those motherboards to enable SLi code in nVidia's drivers. nVidia makes considerably less money selling magic words than chips. But without SLi support, who is going to buy multiple high end nVidia cards?

The biggest "surprise, didn't see that coming" market for high end video cards is Wall Street who are using them to build super computer trading systems that buy stock on news milliseconds before the competition and then sells to that competition for an immediate gain seconds later for instant profit.

Rumor is nVidia may have a video card with their G300 GPU to show before the end of this year. Nothing about when you will be able to buy one but it sounds like the first batch of G300's are allocated for Wall Street and not die hard gamers. The new price point for high end video cards as set by ATI's HD 5870 is discouraging nVidia's manufacturers from building the top end cards because the fat profit margin there is quite thin now. So it's been getting tougher to find GTX 275/285/295 cards from big name players other than eVGA for some reason, I'm guessing price, eVGA really haven't cut theirs.

ATI has hit a snag with their 5xxx series cards, two actually. First they didn't plan they will be so popular so they didn't order enough 40nm wafers the chips are made of and second the Fabs producing them are having manufacturing problems at 40nm so yields are horrible making the shortage even worse. So it's tough if not impossible to find the HD 58xx series cards. The GPUs for the HD 57xx series don't seem to be in short supply but they are literally half of an HD 58xx GPU in size and complexity.

Basically it sucks now if you are looking at top shelf video cards.

It's also tough to find a mid to high end video card that isn't dual slot, explicitly (double wide bracket) or practically (single width bracket but huge bolted on heatsink interfering with the next slot. Problem is simple. The more powerful the GPU, the more power it uses and the more heat it produces. For an air cooled solution the more cooling fin area you have the better. Also more area generally means a less noisy/slower fan can provide cooling. Add this all together and poof, double wide video cards starting at $100-125.

Multiple GPU support is not automatic in games. However games aren't coded to know if multiple GPUs are in the system but they do need to be coded so the CPU has time to kill between the last command it issued to the GPU and the GPU finish rendering the frame. Why, because if you have a second, third, etc. GPU in the system, the CPU can use that time to start issuing commands for the next frame to the next GPU in sequence. It also helps if you are pushing the GPU with crazy sick shaders, a large resolution (lots of pixels) as well as multisampled anti-alias and anisotropic filters to maximize the time it takes the GPU to finish the frame. The more the CPU waits on a single GPU system, the more a multi-GPU system will help framerate.

This is also why two lower level video cards show such a performance boost in multi-GPU setups. Or when the CPU is much, much faster than the GPU. Again, the longer the CPU needs to wait for the GPU to finish, the more multiple GPUs help.

This is why it's not an automatic "home run". It depends on the game. For our game I'm not sure if the GPU is simply not pushed hard or the engine is coded to command the GPU bits at a time leaving little time between the last command and the frame finishing.

However I'm going out on a limb here and say that full volumetric shadows is an expensive post process effect along with more realistic reflection in buildings and water. I think it will be safe to say that those two effects are going to load a ton of work on to the GPU at the end of each frame. Because of this I'm guessing we will see a more obvious improvement in framerates on multi-GPU rigs in the Going Rogue expansion than we do today. How much? I can't say.

Father Xmas: been pushing the developers for an official comment on what the stance of accelerated physics development is, if they intend to remain with PhysX or swap it out. So far... no official comment

*pulls out the 15ft poking stick and tries to get within range of a Wild Positron*