8800GT pricing...

....isn't this backwards compatible - stick it into normal pcie 1 ?
 
No 8800GT is PCI-E 2.0 one of the highlights of the board that NVIDIA is so excited about...
 
Honestly I think this pci express 2 and pci express 3 is a money making scheme! I just read the wikipedia article on pci express and the part that describes pci express 3 says that the only difference between 2 vs 3 is the encoding scheme which the engineers "discovered" if they dropped it, the speed effectively ramps up from 5gb to 8gb.

As for the version 1 vs version 2, it's fully backward compatible but it has more bandwidth (which you won't realize the potential of).
 
No that's not necessarily true. Just because you are unaware of the benefits iy doesn't mean they are non existent :(

Depending on the GFX card and its clocks, bumping up PCI-E speed from reference 100MHz even to 110MHz gives a speed bump in frame rates and that's on 16X PCI-E lane... So obviously bandwidth is still a limitation with high end cards that can tolrate more than the PEG 1.0a standard provides

Right now for instance, the 8800GT overclocked to 702/1728/2000 Is beating the 8800Ultra overclocked to 702/1728/2430 in 3DMark05. Naturally the 8800GT isn't as strong as the Ultra, but it does get over twice the interface bandwidth when compared to the 8800Ultra...

Not saying this is the only reason it's faster but it sure does help...
I haven't tried the 8800GT on a PCI-E 1.0a board though (at least one I can use the same components on to compare)
 
As far as I know if you overclock the bus you are also overclocking the videocard.

I don't think the videocard generates it's own clock frequency (and I haven't see any Quartz Crystals on them either), so that would mean it generates it's own clock signal by either a clock multiplier or clock divider.

Anyway only way to know for sure is to benchmark a card in both modes.

But I still seriously doubt the videocard needs *that* much bandwidth, video encoding is different because you deal with uncompressed data go back and forth but games usually have compressed textures.
 
i thought that the cards today dont fully use all the bandwidth on the bus?

its a money making scheme just like ddr3 :)

such a small performance difference


shock u have 8800gt on a pci 2 board
im sure you have a pci 1 board

how about benching both for us and showing us whats the difference is
 
I c a few bargains coming my way with all the 8800gts's thats gonna get sold:D. PCI-E 1 give something like 75watt on the slot and PCI-E 2 somthing like 160 watts I think so just in raw power supply on the slot the card can run better (thats way all the top cards need external power source).
 
Lol what i find funny is all the lads that have listed 8800GTS's just recently for sale in the classifieds at myadsl give the reason for sale is to upgrade to a GTX lmao

That has GT written all over it
 
haha, yup killa. :p

At least his price is good.

Shock, if you could do a review on PCI-e 1 vs 2, afaik you'd be the first person with one out. :D Only problem is you'd have to change to a completely different mobo, and the performance difference should already be very small so that might just destroy it. No way to force PCIE 1 mode on the board/BIOS?
 
///MoH said:
Lol what i find funny is all the lads that have listed 8800GTS's just recently for sale in the classifieds at myadsl give the reason for sale is to upgrade to a GTX lmao

That has GT written all over it
Click to expand...

yep, my observation aswell. The one on another forum is especially funny. The person sells it due to "not using the capabilities" so gonna downgrade (That is in the classifieds) Then i found a thread of that same person saying he ordered the GT :D
 
haha Gnome I think you got it slightly wrong there...
Overclocking the PCI-E bus has nothing to do with GFX card clocks...
PCI-E bus (primary at least) is located in the Northbridge with the MCH however because it's a system interface bus like SATA/USB etc... it's clock is controlled by the same PLL that controls the South bridge where all these are located. These however are locked to certain dividers or frequencies. 33MHz for PCi (SATA etc..)/100MHz for PCI-e so you can clock FSB(Northbridge) without affecting the PCI clocks etc...
PCI-E clock governs data rates for the interface not the graphics card.
GFX cards have their own PLL that determine clock speeds 2 at the least... One for memory, one for gfx core.

Textures are not the only thing that go to the video card, there are other things like object data, texture id, vertex data etc... Also gfx cards today can't do DMA, so they have to go through NB/MCH and the only way to go through that is through the interface. Overclockers all over use PCI-E bus sometimes up to 159MHz to boost gfx card performance, and there is a sizable benefit believe it or not. ;)
Heck use it myself even :P

It would have been nice to be able to test, but X38 chipset is set at 2x16 lane PCI-E so there's no 8X mode at all... What is evident however is that with e 8800GT oc PCI-E make zero difference because interface isn't saturated yet, but I suspect with future gfx cards it may eventually make a difference once again.
 
I read that a 8800GTX cant saturate a full 16x PCIe 1 interface.. not only that it cant even saturate a 8x PCIe 1 interface.. no way in hell the 8800GT is gonna do it..Hence i doubt there will be a performance difference between PCIe 1 and 2

Cant find the linkys... if i find them will put them up
 
ShockG said:
haha Gnome I think you got it slightly wrong there...
Overclocking the PCI-E bus has nothing to do with GFX card clocks...
PCI-E bus (primary at least) is located in the Northbridge with the MCH however because it's a system interface bus like SATA/USB etc... it's clock is controlled by the same PLL that controls the South bridge where all these are located. These however are locked to certain dividers or frequencies. 33MHz for PCi (SATA etc..)/100MHz for PCI-e so you can clock FSB(Northbridge) without affecting the PCI clocks etc...
PCI-E clock governs data rates for the interface not the graphics card.
GFX cards have their own PLL that determine clock speeds 2 at the least... One for memory, one for gfx core.
Click to expand...

I think you might have misunderstood what I was saying earlier, what I was saying is that I thought the Gfx card clock signal is multiplied from the PCI-Express clock but I tested this theory at it is not.

Anyway the reason I assumed it might be so is, I did a course on Digital Circuits and another on Microprocessors at varsity for extra credit where we built circuits with 4000 series IC's and Microchip PIC microcontrollers and in short we were told to multiply or divide a clock frequency and to avoid having more than one clock generator active (the reasons are beyond the scope of my thread but I'll be happy to post why with evidence from my oscilloscope).

Anyway the point of this drivel is that I though a computer motherboard would work in the same way, IE all clock signals are divided/multiplied from a single clock. Anyway the fact that the Gfx clock doesn't change when I overclock the PCI-Express from 100Mhz to 110Mhz proves to me that the clock signal is either A) generated on the videocard itself or is B) Generated by a very complex clock multiplier that takes the input frequency in consideration, it could also perhaps be that one of the PCI-Express lanes only send a reference clock signal which is then used by the frequency multiplier/divider.

Regardless, I now know that overclock the Pci-Express bus DOES not overclock the Gfx card, my bad. Although I still question whetever you'll see benefit from such a overclock ;) Benchies from a reputable site would sway my opinion tho (or you can do it yourself if you want), Tomshardware did some articles where they tested it and it had no influence, but it's quite old, would be interesting if it's change tho...

EDIT: Also the PCI-Express 2 vs 3 is still a money making scheme, even if it increases performance :D The reason I say that is if you build a digital circuit you only add abstraction or make the circuit more complex if you absolutely have to, for Pci-Express 3 they remove a layer of abstraction to improve performance, but why have this layer in the first place if it where never needed? :) Anyway only they would know...
 
Last edited:
I read that a 8800GTX cant saturate a full 16x PCIe 1 interface.. not only that it cant even saturate a 8x PCIe 1 interface.. no way in hell the 8800GT is gonna do it..Hence i doubt there will be a performance difference between PCIe 1 and 2
Click to expand...
not necessarily, you can't make that assumption, because allot has changed in the G92 as opposed to the G80. IN a gfx core, there could be anything from 400 or more pipeline stages, where the stall was in the G80, it isn't there in the G92, maybe moved further on towards the render output or further back towards geometry setup etc... We can't know where the PCI-E 2.0 bandwidth will relieve this stall... :/

Also the tests done, remember were not on CPU's as powerful as we have now, RAM wasn't as fast and the graphics card drivers were not as refined as they are now. You notice this when the gfx cards have been overclocked and the interface starts to become the bottleneck to some degree...

I have seen this myself. Any overclocker will tell you that PCI-E speed makes a difference, you won't see it on Anandtech and the like because they are not usually reviewing on 4GHz+ machines etc.. :/

Try it yourself even. load up 3DMark2001se, Game Test4 (Nature) run it and record the frame rate. Restart the PC and change PCI-E frequency to 112MHz and see what happens. ;)
 
ShockG said:
Try it yourself even. load up 3DMark2001se, Game Test4 (Nature) run it and record the frame rate. Restart the PC and change PCI-E frequency to 112MHz and see what happens. ;)
Click to expand...

Can't you do it :p (and take screenies?)

JK, I'm busy with exams so I'll try it as soon as they're done :D

EDIT: Btw. is it only on 3DMark or on real games like say... Bioshock, because Nature has huge textures....
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X