Is 9800gtx worth investing in PCI-e

slapnutz

Regular
Ok, if was looking to build a pc with upgarding to a nVidia 9800xxx (or whatever it'll be called), is their any point in getting a PCI-E 2.0 capable motherboard?

Would that card exploit the faster bandwidth of Pci-e 2.0?... or... does that depend if the software/game actually pushes the card that far?

Is there any senarios in the current age that uses x16 PCI-e 1.0?

Sorry if it seems a bit noobish.
 
Last edited by a moderator:
I don't think we know yet. The 3870X2 uses a PCIE-1.x bridge, so it doesn't have much way to benefit from the PCIE 2.0 interface. But I don't recall hearing anything on the 9800??? card's interface controller...

Now, as for this:
Is there any senarios in the current age that uses x16 PCI-e 1.0?

Perhaps the even more basic question is: does PCIE 2.0 even matter?

I see a lot of PC veterans making the claim that "current games don't use PCIE 1.1 16x even right now", which is a complete falsity if you think about it. You don't have to "saturate" a bus in order to gain from a higher transfer rate. When was the last time you "saturated" your SATA 150 connection to your local hard drive? Why did we all move from ATA/133; we weren't "saturating" it either? Hell, consumer-level hard drives at just now starting to hit above 100mb/sec of actual sustained throughput capability, and it's how many years after ATA-133 standard came out?

The problem is, people associate "saturation" with minutes and minutes and hours and days of 100% usage. The reality is, you more than likely "saturate" your PCI-E 1.1 bus when your machine is loading all the game textures into VRAM. You more than likely "saturate" your PCI-E 1.1 bus when you CTRL-TAB switch from a video game. You're getting benefit out of the higher bandwidth, but you're only seeing it when the card doesn't have the needed resources within its' own VRAM.

The reality is, PCIE 2.0 matters even now, but it's really only visible when your video card is under so much stress that the entire gammut of performance is in the toilet. If you have so much video data that it cannot all fit within the local video memory, you will instantaneously see the benefit (or detriment) of your underlying bus topology. The problem then becomes this: your video subsystem is so far against the wall, you will still complain about the performance -- even if it IS ~2x faster than PCI-E 1.x

It's not a difficult thing to understand; it takes only a few minutes to realize that the difference between waiting on the bus versus waiting on the interconnected devices.
 
Last edited by a moderator:
Since most of them will come from HD and not system memory even that scenario isn't very common.

Very true. I was kinda leaning towards that whole Vista superfetch thing when I was writing that, as in, hoping that the data was already cached from the harddrive when you launched the app. I've actually seen Vista do this with the games I play the most...

That, or maybe a case where you need to squeeze in a bit more texture information on a "big" scene change that the engine has already cached for you. Dunno, guess it just depends...
 
Yeah one thing is for sure though, if you're card has enough vram (i.e you're not going to be laoding textures from system ram) there's going to be no performance increase moving from PCIe 1.0 to PCIe 2.0. Heck I highly doubt there would be any difference between a 8x lane and a 32x lane with any any end card.
 
Wait a second: Wouldn't you use the bandwidth when you Alt-tab between windows and a game? the information has to be cached in main memory.
 
Oh yeah you would, but that's nowhere close to a "majority of the time" exercise. I mentioned it above:
The reality is, PCIE 2.0 matters even now, but it's really only visible when your video card is under so much stress that the entire gammut of performance is in the toilet. If you have so much video data that it cannot all fit within the local video memory, you will instantaneously see the benefit (or detriment) of your underlying bus topology. The problem then becomes this: your video subsystem is so far against the wall, you will still complain about the performance -- even if it IS ~2x faster than PCI-E 1.x
 
Odd that the PCIE 2.0 supposedly had a leg-up at every setting; I have to wonder what else might be going on. Some of those scenarios do make sense though, nice to see some numbers put to it.

The moral of the story is: don't skimp on video memory :)
 
Back
Top