PCI-E 2

S

slugger_2003

Limp Gawd
Joined
Jan 30, 2007
Messages
365
How soon does everyone think it will be till PCI-E 2 will actually be utilised? Of course nobody actually knows yet as nothing out there now which is commercially available does (as far as I know), I'm just interested as to whether or not it's worth adding the consideration of PCI-E 2 when looking for a new motherboard.

All input appreciated
 
I made a thread asking if the 4870 takes advantage of pci-e 2, answer seems to be no, so now I'm asking for peoples opinions on when pci-e 2 will be utilised.
Slightly difference topics, hence different thread.
 
OK, the answer is that the lower end cards will probably be the only ones to take advantage of it any time soon. So unless your budget hunting cards in the future or want that last few FPS in SLI or Crossfire it isn't going to make a big difference.
 
Lower end cards will take advantage of pci-e 2 before higher end cards will? That just doesn't make sense to me. Surely higher end cards send a higher bandwidth of data, hence the need for a larger interface? Why would cards sending less data require a larger interface? To me that's like saying you would need an olympic powerlifter to carry a bag of flour and a 12 year old geek to lift a 500lb weight...haha, perhaps not the best analogy but I think it gets my point/confusion across?
 
slugger_2003 said:
Lower end cards will take advantage of pci-e 2 before higher end cards will? That just doesn't make sense to me. Surely higher end cards send a higher bandwidth of data, hence the need for a larger interface? Why would cards sending less data require a larger interface? To me that's like saying you would need an olympic powerlifter to carry a bag of flour and a 12 year old geek to lift a 500lb weight...haha, perhaps not the best analogy but I think it gets my point/confusion across?
Click to expand...

your not see this in the proper terms is all, for a high end card the PCI bus only needs to deliver the frames to the GPU fast enough. everything else is handled on the card. that is why on high end systems the CPU is by and large of little difference, or the PCI bus after a point. You only need so much because the card can't render it any faster, often there is no point in do so anyways (most LCD screens only refresh 60 times a second, so FPS over that is wasted.) On a low end card with limited memory its a different story. your having to page more things from the system memory. In some cases the CPU is now taking more of the load as well. that is why the lower end does better on the higher bandwidth bus then the high end cards.
 
slugger_2003 said:
Lower end cards will take advantage of pci-e 2 before higher end cards will? That just doesn't make sense to me.
Click to expand...

He's got a point there. Low end cards with smaller memory sizes (or IGPs with no dedicated memory) would benefit from the faster bus because they have to fetch more textures from main memory whereas high end cards can just prefetch everything and keep it all in fast onboard ram.

Going from 5 to 8 FPS is technically a huge improvement ;)
 
Valset said:
your not see this in the proper terms is all, for a high end card the PCI bus only needs to deliver the frames to the GPU fast enough. everything else is handled on the card. that is why on high end systems the CPU is by and large of little difference, or the PCI bus after a point. You only need so much because the card can't render it any faster, often there is no point in do so anyways (most LCD screens only refresh 60 times a second, so FPS over that is wasted.) On a low end card with limited memory its a different story. your having to page more things from the system memory. In some cases the CPU is now taking more of the load as well. that is why the lower end does better on the higher bandwidth bus then the high end cards.
Click to expand...

What a bunch of crap. Like you don't need interfacing with the cpu or system memory? The card and GPU still need to interface with the rest of the system at a very high rate of speed. Typically, the higher the res and the more complicated the frame - the higher bandwidth that is needed, not the other way around. The gpu JUST draws the frame - the cpu figures out WHAT to draw.
http://www.tomshardware.com/reviews/pci-express-scaling-analysis,1572-8.html
 
Mega666 said:
What a bunch of crap. Like you don't need interfacing with the cpu or system memory? The card and GPU still need to interface with the rest of the system at a very high rate of speed. Typically, the higher the res and the more complicated the frame - the higher bandwidth that is needed, not the other way around. The gpu JUST draws the frame - the cpu figures out WHAT to draw.
http://www.tomshardware.com/reviews/pci-express-scaling-analysis,1572-8.html
Click to expand...

hum, how does one get both feet in his mouth? what that articale actually shows is what I mentioned above, that the speed of the PCIe bus stops mattering much after a point. and you need to brush up on how a graphics card works, the higher the resolution and details the more GPU bound you become, not CPU bound

http://www.guru3d.com/article/cpu-scaling-in-games-with-quad-core-processors/
explains how CPU scaling works. please not that the frames per second only differ slightly over the range of CPU. COD4 is a good example

http://www.tweaktown.com/articles/1472/1/intel_p45_vs_x48_crossfire_performance/index.html
not much difference in frame rates from 8X by 8X vx 16X by 16X despite doubling the speed.

damn I am bored tonight.
 
Would really like to know if the 4870x2 is going to need 8x 8x or 16 x 16x p45/x48/x38 board over a p35. Guess we'll all find out.
 
Valset said:
your not see this in the proper terms is all, for a high end card the PCI bus only needs to deliver the frames to the GPU fast enough. everything else is handled on the card. that is why on high end systems the CPU is by and large of little difference, or the PCI bus after a point. You only need so much because the card can't render it any faster, often there is no point in do so anyways (most LCD screens only refresh 60 times a second, so FPS over that is wasted.) On a low end card with limited memory its a different story. your having to page more things from the system memory. In some cases the CPU is now taking more of the load as well. that is why the lower end does better on the higher bandwidth bus then the high end cards.
Click to expand...

So you're saying we should be running the HD4870X2 on a normal PCI bus. In fact, why not go back to ISA while we're at it.

LOL @ you!
 
Valset said:
hum, how does one get both feet in his mouth? what that articale actually shows is what I mentioned above, that the speed of the PCIe bus stops mattering much after a point. and you need to brush up on how a graphics card works, the higher the resolution and details the more GPU bound you become, not CPU bound

http://www.guru3d.com/article/cpu-scaling-in-games-with-quad-core-processors/
explains how CPU scaling works. please not that the frames per second only differ slightly over the range of CPU. COD4 is a good example

http://www.tweaktown.com/articles/1472/1/intel_p45_vs_x48_crossfire_performance/index.html
not much difference in frame rates from 8X by 8X vx 16X by 16X despite doubling the speed.

damn I am bored tonight.
Click to expand...

Bored? Well try and learn something about graphics cards and computers. Is it just me or has this forum filled up with posts of idiocy lately?
 
Valset said:
your not see this in the proper terms is all, for a high end card the PCI bus only needs to deliver the frames to the GPU fast enough. everything else is handled on the card. that is why on high end systems the CPU is by and large of little difference, or the PCI bus after a point. You only need so much because the card can't render it any faster, often there is no point in do so anyways (most LCD screens only refresh 60 times a second, so FPS over that is wasted.) On a low end card with limited memory its a different story. your having to page more things from the system memory. In some cases the CPU is now taking more of the load as well. that is why the lower end does better on the higher bandwidth bus then the high end cards.
Click to expand...

But if a low end graphics card only sends 1 unit of data (the unit is a fixed size) and the PCI-e 16x bus can handle 2 units of data then the 16x bus is only being half utilised, correct? What use would increasing the bandwidth have here? It's like the someone buying 8GB of memory just for simple tasks like word processing and watching the odd film, the full amount of memory is never used so it's absolutely pointless.
Perhaps the step up from pci-e 16x to pci-e 2.0 is actually the speed the data is sent rather than the amount of data that can be sent in a single clock cycle? If so I understand the concept here. I read a discussion about gddr5 256-bit bus can be compared to gddr3 512-bit bus, it seemed to be stating a balance between both speed and amount of memory.

Perhaps I should have been clearer about what I understood bandwidth to be in this circumstance. So PCI-e 2.0 has the same bus size as PCI-e 16x but just sends data at a higher frequency?
 
slugger_2003 said:
Perhaps I should have been clearer about what I understood bandwidth to be in this circumstance. So PCI-e 2.0 has the same bus size as PCI-e 16x but just sends data at a higher frequency?
Click to expand...

You are confusing your terms. The 1.0/1.1 and 2.0 refer to the revision, the x16 refers to the lanes.

PCI-E 2.0 transfers twice the data per lane as PCI-E 1.0/1.1

Therefore, PCI-E 2.0 x16 is twice as fast as PCI-E 1.0/1.1 x16. Also, PCI-E 2.0 x8 is the same speed as PCI-E 1.0/1.1 x16.

So just saying PCI-E 2.0 or PCI-E x16 doesn't mean much, as in the former case you aren't stating the number of lanes, and in the latter you aren't stating the revision.
 
You must log in or register to reply here.
Back
Top