Surprise Surprise Pci Express does next to nothing

[Corporate Thug]

http://www.hothardware.com/viewarticle.cfm?articleid=542

 

[LTC8K6]

Yet.

Sounds normal for new stuff.

 

[AristoV300]

Yeah it just like the Prescott being underclocked so that it's long pipelines hurting and the lack of SSE3 apps. I am sure later in the year when all the newer hardware starts coming out and the Tejas then it may show some improvement. That is why ATI has said it will not stop making AGP cards for later models any time soon.

 

[AristoV300]

Yeah it just like the Prescott being underclocked so that it's long pipelines hurting and the lack of SSE3 apps. I am sure later in the year when all the newer hardware starts coming out and the Tejas then it may show some improvement. That is why ATI has said it will not stop making AGP cards for later models any time soon.

 

[JBT]

no news here. It kinda just like 4 x and 8 x AGP doing next to nothing for graphics cards as well.

 

[Falloutboy]

Originally posted by: AristoV300
Yeah it just like the Prescott being underclocked so that it's long pipelines hurting and the lack of SSE3 apps. I am sure later in the year when all the newer hardware starts coming out and the Tejas then it may show some improvement. That is why ATI has said it will not stop making AGP cards for later models any time soon.

sorry to tell ou but tejas has been scraped

thier going to dothan based chips after prescott maxes out

 

[StrangerGuy]

No big surprise that the PCI-E doesn't give a single drop of performance increase.

 

[nitromullet]

...but ...but, the marketing guys said...

 

[ZobarStyl]

And after all that anger over the AGP to PCI-E bridge, it doesn't do a single thing to performance...lol
So much for anyone who wanted to early adopt PCI-E for any other reason than dual-card setups...I'll be waiting till at least second gen PCI-E for my switch.
Quick question...is AMD even planning on supporting PCI-E any time soon after all this? I know they are waiting on DDRII, maybe they'll just let Intel burn its money blazing a (currently) fruitless path on PCI-E too...

 

[Falloutboy]

Originally posted by: ZobarStyl
And after all that anger over the AGP to PCI-E bridge, it doesn't do a single thing to performance...lol
So much for anyone who wanted to early adopt PCI-E for any other reason than dual-card setups...I'll be waiting till at least second gen PCI-E for my switch.
Quick question...is AMD even planning on supporting PCI-E any time soon after all this? I know they are waiting on DDRII, maybe they'll just let Intel burn its money blazing a (currently) fruitless path on PCI-E too...

I beleave its a chipset thing so as long as the chipsets support it.

 

[BFG10K]

Surprise Surprise Pci Express does next to nothing

AGP does next to nothing.
DDR does next to nothing.
100 BaseT does next to nothing.

Those were the same comments coming out when the technology first came out. And now look where we are today.

 

[MDE]

Originally posted by: BFG10K

Surprise Surprise Pci Express does next to nothing

AGP does next to nothing.
DDR does next to nothing.
100 BaseT does next to nothing.

Those were the same comments coming out when the technology first came out. And now look where we are today.

Yeah, but AGP 8x still has no tangible benefits over 4x so why bother with an even faster solution?

 

[Torghn]

Originally posted by: MDE

Originally posted by: BFG10K

Surprise Surprise Pci Express does next to nothing

AGP does next to nothing.
DDR does next to nothing.
100 BaseT does next to nothing.

Those were the same comments coming out when the technology first came out. And now look where we are today.

Yeah, but AGP 8x still has no tangible benefits over 4x so why bother with an even faster solution?

The move isn't to add more speed, it's to move to a better and more advanced technology (scalable). Since the change is being made why not make it faster? It may not help now, but it will later, and it sure doesn't hurt anything.

Another reason PCI-express is better is that you can easily run 2 high end video cards, just like the Voodoo 2s use to. This sort of setup isn't possible with AGP.

 

[bunnyfubbles]

Originally posted by: MDE

Originally posted by: BFG10K

Surprise Surprise Pci Express does next to nothing

AGP does next to nothing.
DDR does next to nothing.
100 BaseT does next to nothing.

Those were the same comments coming out when the technology first came out. And now look where we are today.

Yeah, but AGP 8x still has no tangible benefits over 4x so why bother with an even faster solution?

yes, let's stop advances in hardware developement and stick with what we've got until the end of time.

 

[hysperion]

sounds like it may be possible but I don't know about easy....

 

[caz67]

I agree , that the improvement at this point is neglible to say the least. However i think the technology behind the new chipsets is impressive.

I felt that the industry has gone stale, i think it was a great move for the future. It will take time for every thing to mature.

 

[DerwenArtos12]

You really gotta think about the drivers that are probably out right now too, I would bet they are barely alpha drivers.

 

[Corporate Thug]

Originally posted by: bunnyfubbles

Originally posted by: MDE

Originally posted by: BFG10K

Surprise Surprise Pci Express does next to nothing

AGP does next to nothing.
DDR does next to nothing.
100 BaseT does next to nothing.

Those were the same comments coming out when the technology first came out. And now look where we are today.

Yeah, but AGP 8x still has no tangible benefits over 4x so why bother with an even faster solution?

yes, let's stop advances in hardware developement and stick with what we've got until the end of time.

advance implies some sort of gain or increase...

if advances are going to be things like going from AGP 4X to 8X, i can live without it

 

[PrayForDeath]

Originally posted by: shady06

Originally posted by: bunnyfubbles

Originally posted by: MDE

Originally posted by: BFG10K

Surprise Surprise Pci Express does next to nothing

AGP does next to nothing.
DDR does next to nothing.
100 BaseT does next to nothing.

Those were the same comments coming out when the technology first came out. And now look where we are today.

Yeah, but AGP 8x still has no tangible benefits over 4x so why bother with an even faster solution?

yes, let's stop advances in hardware developement and stick with what we've got until the end of time.

advance implies some sort of gain or increase...

if advances are going to be things like going from AGP 4X to 8X, i can live without it

PCI-Express is not meant to give us instant increase in performance, check out what Dave Orton said about it:

HWZ : Looking at PCI Express, some manufacturers we spoke to were quite skeptical of its performance benefits as compared to AGP. How confident is ATI in delivering the kind of performance envelope that power users require?
Dave: The foundations we have to remember about PCI Express is that AGP is really pretty much tapped out. In order to get to the next level of performance, we really have to double the bus width. So what PCI Express did was came in and cut the AGP bus width in half because ultimately what drives power is pin-out and cost is driven by pin-out. So what we've been able to do is reduce the bus width and as you can see, it almost doubles the performance from a bandwidth standpoint. So in terms of immediate performance gains, I don't think that was the primary objective of PCI Express. The primary objective was to create a foundation for the next generation of visual platforms. And you have be able to go in and in fact, re-architect at the platform level, recognize that you're not going to see instantaneous performance advantages when you go in and affect double infrastructure. It's like the network infrastructure, you might run it ten times faster, not today, but it gives you the headroom and growth opportunity. So I think PCI Express is really there to create a new foundation for us to grow from because AGP was tapping out. Now we have specific applications where you can go in and see performance advantages. The generic applications today, if they are not pushing AGP to the limits, they are not going to push PCI Express to the limits either. It's really the next generation of development that's going to do that.
HWZ: In your opinion, how quickly do you think the market will switch over to PCIe? The switch from PCI to AGP was a pretty smooth transition, do you think we'll see the same with AGP 8x and PCIe?
Dave: What drives the switching of the industry is (to look at) what's being provided by the OEMs and I think Intel is providing Grantsdale as the chipset of choice, [and] moving forward to provide next generation processors and next generation speeds of front side buses. ATI, and you'll see over time with NVIDIA, our next generation of graphics solutions [running] natively on PCI Express. And so if the industry wants to move forward technologically, it's going to have to move forward with PCI Express. I think the whole PC industry thrives on this idea of better, faster and cheaper. PCI Express is going to prove to be a better interface over time. The OEMs are going to switch very fast. I think the channel market will take a little longer to switch. We think of it as a 12 month transition but it's going to go fairly quickly in the OEMs and I think you'll see the high-end performance segments go fairly fast in the channels as well.
HWZ: So, if you have both AGP and PCIe VPUs, isn't it very difficult to control the number of SKUs you carry, and isn't it a risky business to build so many different chips as well?
Dave: Gee, would you rather have one SKU and no business, or five SKUs and lots of business? That's an easy answer for me. (laughs) So ultimately you've got to solve the customer's problems and so for us, in the low-end, it's all about cost and capability. In the high-end, it's about performance. And our view is that having a single SKU and a stack of AGP parts doesn't solve either of those problems on this side of the fence. So it's just not the right long term solution. We actually taped out our first native PCI Express solution twelve months ago. That was the demonstration vehicle which ultimately now is the X600. The first generation of that was not even a low-k process but it was validating both the Intel Grantsdale solution and our design. We went through a couple of iterations of that, not just ATI, but ATI and Intel went through iterations of both chips to be able to optimize that. So when we finally entered calendar year 2004, we knew we had a solid interface and now we deployed that in 0.11µm and into our high-end X800 solution. It was a solid base interface we're working with. So, you always have to take some risk in this industry if you want to innovate and we decided to innovate.
Stanley: Risk is an inherent part of this business and we accept a lot of risk and we execute them well in order to minimize the impact of risk. The reason why there's only two major graphics players, excluding Intel, is that it's not easy to support all the OEM customers and all the risks and all the developments associated with the software and hardware. It's a challenging business to be in and we managed to do it very well. So, risk is something that Dave is good at dealing. (laughs)
Dave: That's why I don't have any hair back here. (laughs)

You can read the full interview here

 

[biostud]

Originally posted by: Torghn

Originally posted by: MDE

Originally posted by: BFG10K

Surprise Surprise Pci Express does next to nothing

AGP does next to nothing.
DDR does next to nothing.
100 BaseT does next to nothing.

Those were the same comments coming out when the technology first came out. And now look where we are today.

Yeah, but AGP 8x still has no tangible benefits over 4x so why bother with an even faster solution?

The move isn't to add more speed, it's to move to a better and more advanced technology (scalable). Since the change is being made why not make it faster? It may not help now, but it will later, and it sure doesn't hurt anything.

Another reason PCI-express is better is that you can easily run 2 high end video cards, just like the Voodoo 2s use to. This sort of setup isn't possible with AGP.

But only in special builds that actually have two 16x PCI-e slots, like alienwares ALX.

 

[ZobarStyl]

I hope that some non-Alienware non-server level boards have the dual x16 slot config...it seems much more practical since dual PCI-E wouldn't hamper performance at all, whereas 2 AGP cards in one mobo might actually come close to using all the bandwidth, and thereby take a performance hit. It just seems a shame to give us all this bandwidth and not let us use it to it's fullest...

 

[JBT]

The good thing with PCI-E is that is can supply more power to the video cards correct? I don't see that as a bad thing at all. I agree that it is good to move forward and often the first generation does next to nothing in terms of performance but as time goes by we can start to take advantage of PCI-e's benifits. What I meant by my post was I think everyone here already knew or should have know by now that PCI-e wasn't going to be offereing much in terms of performance this time around.

 

[Cat]

One huge PCI-E advantage over AGP is asynchronous data transfer. You don't have to stall the entire graphics pipeline to read data back from the GPU. This means MUCH faster framebuffer reads, render-to-texture, occlusion culling, and other useful things.

Once applications make heavier use of these techniques, we'll see the benefits of PCI-E, I hope.

 

[kylebisme]

i can't say exactly what pci-express will or will not do as i have never seen it in action; but i hardly think that low res, cpu limited, ut2004 botmatchs would be the way to test the effects of the new bus system.

 

[reever]

Well there is one quantifiable benefit

http://techreport.com/reviews/2004q2/intel-9xx/index.x?pg=13

But what impact this speed has on anything is unknown