ATI to bridge their chips

[Keysplayr]

ATI to bridge their chips after all.
So much for brow-beating Nvidia for doing it.

 

[ss284]

There is a slight difference, seeing as ati is going from pci-e to agp, meaning the card is natively pci-e, unlike nvidia which is the exact opposite, a natively agp card with an agp to pci-e bridge chip. Meaning the performance impact from the bridge chip (no matter how small) will be experienced by the pci-e versions of nvidia's cards and the agp versions of ati's cards.

-Steve

 

[Keysplayr]

Originally posted by: ss284
There is a slight difference, seeing as ati is going from pci-e to agp, meaning the card is natively pci-e, unlike nvidia which is the exact opposite, a natively agp card with an agp to pci-e bridge chip. Meaning the performance impact from the bridge chip (no matter how small) will be experienced by the pci-e versions of nvidia's cards and the agp versions of ati's cards.

-Steve

And why is this significant?

 

[darkswordsman17]

Its kinda funny don't ya think.

ATi criticized nVidia so badly over their optimizations, and lets see, looks like ATi was doing it as well.

Then they brag about how their cards will be better because they won't be using a bridge solution, and lets see, looks like they will after all.

I think ATi might wanna start watching themselves when they criticize nVidia. They're lucky they haven't caught more flak over this. Don't get me wrong its not like it makes nVidia any better.

Both companies should just stop talking sh!t on each other unless they can thoroughly back it up.

And why is this significant?

Its not significant. Like he said there is a slight difference. ATi's bridge should in theory be better because its starting with the faster interface and converting to the slower one, while nVidia's is converting from the slower to the faster. I don't now if it'll make much difference or not, I doubt it will, but who knows.

 

[kylebisme]

it is a long story keys. well actually it isn't really a long story, but i don't expect you to understand anyway. just keep thinking that ati is a bunch of two-faced liers who talk smack only to turn around and do the same thing.

 

[jasonja]

Because PCI-E is faster so bridging it to a slower AGP bus should NOT hurt performance since the bus is the limiting factor. nVidia's approach makes the card/bridge the bottleneck since it's on a faster bus.

 

[nitromullet]

Well I have to agree that this stlightly different. ATi, is using the bridge to support legacy hardfware. The native solution is the newer, superior tech. Also, we all know that ATi can successfully implement an AGP based card, so bridging back 'down' to AGP from PCI-E isn't as big of a deal as bridging 'up' from AGP to PCI-E when you don't have a natively PCI-E solution.

 

[jtelles]

From what I got from the story was that ATi was going to bridge their AGP cards to PCIE now, then in the future, when PCIE is more avail, they will make native PCIE chips and bridge to AGP then.

 

[Keysplayr]

Ahh, I get it now. If nvidia bridges their chips, its a bottleneck. If ATI bridges their chips, its perfectly fine. All kidding aside, it is very likely that both parties will suffer a bit of performance loss through any type of bridging. Jasonja, even the CEO of ATI states there will be some performance hit. Although he did not mention to what extent.

If I came of as in favor of nvidia Snowman, you would be mistaken. Both companies, bridged or not, kick ARSE. Agreed? I dont think anything either company says or does should be held in so much contempt. Graphics cards are the most cutthroat market out there. Companies do what they have to do to survive. I have a 5900U and its not the fastest card out there, but it suits me just fine until the newer gen comes down in dinero.

 

[darkswordsman17]

From what I got from the story was that ATi was going to bridge their AGP cards to PCIE now, then in the future, when PCIE is more avail, they will make native PCIE chips and bridge to AGP then.

I don't think this is correct, because they are making separate AGP and PCI-E versions of their cards. It sounds like they are going to implement a bridge on just the PCI-E ones until PCI-E becomes more widespread.

Then again, no one will know for sure until they get actual cards in their hands.

The thing I'm wondering is will the people who buy cards with PCI-E native but with the bridge be able to disable the bridge so it runs its native PCI-E. Also, with cards that are AGP native but have a PCI-E bridge be able to run in both?

 

[Insomniak]

Originally posted by: jasonja
Because PCI-E is faster so bridging it to a slower AGP bus should NOT hurt performance since the bus is the limiting factor. nVidia's approach makes the card/bridge the bottleneck since it's on a faster bus.

*rings the retard bell*

BUS SPEED MAKES NO DIFFERENCE!

Games aren't even using the 1.1GB of bandwidth that AGP 4X provides. 8X AGP is useless, and PCIE will be even more useless. It's nothing but a marketing buzz word designed to sell motherboards and video cards to ignorant sheep.

If you think the PCIE standard is going to make ANY differene in performance in the next two years, you're deluding yourself, so let's stop shoveling all this sh*t that a faster bus speed means something, ok?

 

[Lonyo]

Seems like thay are going to be bridging their R420 from AGP to PCI-E NOW, then later, they will bridge the R423 from PCI-E to AGP?
Ie: temp solution until they make their native PCI-E core R423 cards?

 

[VIAN]

AFAIK, they were always gonna do this.

ATI has native PCI-E and they will bridge it to AGP until PCI-E is available.

Nvidia has native AGP and they will bridge it to PCI-E when it is available.

It's not like it matters. Either way there is a bridge and the bus system doesn't really affect framerates----

SO WHY SUCH A BIG DEAL?????????

 

[Matthias99]

Originally posted by: Lonyo
Seems like thay are going to be bridging their R420 from AGP to PCI-E NOW, then later, they will bridge the R423 from PCI-E to AGP?
Ie: temp solution until they make their native PCI-E core R423 cards?

AFAIK:

R420 -- AGP chip, will only be on AGP boards.

R423 -- PCI-E native, used for PCI-E versions of the X800 cards.

R480 -- PCI-E native, may be bridged back to AGP as an R420 refresh?

R5XX -- PCI-E native, probably PCI-E only?

 

[biostud]

Heh I don't understand all the fuss about bridging. The difference will probably be less than 1%.

 

[jasonja]

Originally posted by: Lonyo
Seems like thay are going to be bridging their R420 from AGP to PCI-E NOW, then later, they will bridge the R423 from PCI-E to AGP?
Ie: temp solution until they make their native PCI-E core R423 cards?

No, you're wrong. ATI already has native PCI-E parts. Soon all their parts will only be PCI-E internally. When that happens they will still support AGP systems by adding a bridge that makes PCI-E chips work on the AGP bus. They have no reason to slap PCI-E bridges on their AGP parts since they already have native PCI-E equivalents to sell. This is not the case for nVidia right now. That's the difference, if you don't still understand the difference, then perhaps you should avoid posting.

As for the jackass ringing the retard bell. I'm well aware that AGP isn't a limiting factor, but in my example there's clearly a theoretical difference and bottleneck.

 

[Keysplayr]

Originally posted by: jasonja

Originally posted by: Lonyo
Seems like thay are going to be bridging their R420 from AGP to PCI-E NOW, then later, they will bridge the R423 from PCI-E to AGP?
Ie: temp solution until they make their native PCI-E core R423 cards?

No, you're wrong. ATI already has native PCI-E parts. Soon all their parts will only be PCI-E internally. When that happens they will still support AGP systems by adding a bridge that makes PCI-E chips work on the AGP bus. They have no reason to slap PCI-E bridges on their AGP parts since they already have native PCI-E equivalents to sell. This is not the case for nVidia right now. That's the difference, if you don't still understand the difference, then perhaps you should avoid posting.

As for the jackass ringing the retard bell. I'm well aware that AGP isn't a limiting factor, but in my example there's clearly a theoretical difference and bottleneck.

I think you better take a step back my friend and calm down. If you cant post friendly, then I believe you have no right to tell anyone not to post and should refrain from posting yourself. Take your own advice. As usual, somebody takes a simple topic and goes waaaayy off kilter about it and gets all passionate about it. ITS FRIENDLY CHAT!!!!!!!!!!!! Hello?

 

[jiffylube1024]

Originally posted by: darkswordsman17

From what I got from the story was that ATi was going to bridge their AGP cards to PCIE now, then in the future, when PCIE is more avail, they will make native PCIE chips and bridge to AGP then.

I don't think this is correct, because they are making separate AGP and PCI-E versions of their cards. It sounds like they are going to implement a bridge on just the PCI-E ones until PCI-E becomes more widespread.

Then again, no one will know for sure until they get actual cards in their hands.

The thing I'm wondering is will the people who buy cards with PCI-E native but with the bridge be able to disable the bridge so it runs its native PCI-E. Also, with cards that are AGP native but have a PCI-E bridge be able to run in both?

This is incorrect - AFAK, all of ATI's new cards now are PCI-E; the AGP ones are just bridged.

As a matter of fact, I was just talking to someone who gets cards from ATI (his buddy works there), and he said he can't get any (employee discounted) X800 Pro AGP or X800XT AGP cards - the employees all have access to the PCI-E cards but the AGP ones are so scarce that not even the employees can get their hands on them.

 

[LTC8K6]

I read the story this way: ATI is making two seperate cards now. A native AGP and a native PCI-E.

Later on, when PCI-E has taken over and AGP is relegated to second fiddle, they will bridge the PCI-E cards to AGP for those that still need AGP cards.

Makes perfect sense and is nothing like what NV s doing as far as I am concerned.

 

[jiffylube1024]

Originally posted by: LTC8K6
I read the story this way: ATI is making two seperate cards now. A native AGP and a native PCI-E.

Later on, when PCI-E has taken over and AGP is relegated to second fiddle, they will bridge the PCI-E cards to AGP for those that still need AGP cards.

Makes perfect sense and is nothing like what NV s doing as far as I am concerned.

You may be correct - they may have two flavours now. In that case, it just means that they are producing a lot of PCI-E cards already, which explains the shortages of X800 AGP series cards right now.


Also, you have to put this article in perspective and take it with a whole shaker of salt - this is how the article begins:

JEN HSUN, NVIDIA'S CEO, SAID

You think he's going to say ATI has a smarter solution Nvidia? Not likely!

 

[imported_obsidian]

Bridging really isn't an issue at this time anyways.

 

[Pete]

ATi was deriding nVidia for bridging a GPU optimized for a slower AGP bus to a theoretically faster PEG bus. ATi won't be similarly moving "uphill." (The question that ATi conveniently failed to address was whether nV's GPUs can inherently make use of PEG's speed, or if they'll be limited by their design [cache, etc.] to around AGP speed.)

But this article was about going the other way, from PEG to AGP. Surely you don't expect ATi to maintain two different fab lines for basically the same chip (one for AGP and one for PEG, necessitating more tape-outs and potential problems)? nV will do the same: one chip, and a bridge when nec'y.

Both ATi and nV have said PEG will offer speed increases over AGP, though I'm not sure if that refers to potential increases for games designed around PEG's higher throughput, or current games designed around AGP's handicaps. I guess we'll find out with the first benchmarks how AGP compares to PEG.

I don't see how this is hypocrisy.

 

[ToeJam13]

Because PCI-E is faster so bridging it to a slower AGP bus should NOT hurt performance since the bus is the limiting factor. nVidia's approach makes the card/bridge the bottleneck since it's on a faster bus.

Several sources have stated that nVidia will be using an "internal" AGP spec between their native AGP chips and the PCIe bridge. As such, the inter-communication bus might be as high as 16x (4.2GBps) rather than AGP 8x (2.1GBps).

The joke was that nVidia could do this because their chips had enough headroom to run at higher AGP bus speeds than current x8 specifications allow. ATI's chips couldn't do this, so they were forced to develop a native PCIe solution. Now that they have it, they're rubbing nVidia's nose in the dirt.

Regardless if a video card uses AGP 8x or PCIe, most of the bottleneck will be with the VPU itself. As Insomniak noted, most current video cards aren't even pushing the limits of AGP 8x as it stands. The holy grail of video bus achievements, being fast enough to use system mainboard memory for non-blocking, low latency texture buffers, has not yet arrived.

PCIe needs a killer app to really make it big, like streaming of uncompressed 1080p movies to the video buffer. However, since even compressed 720p/1080i material is hard enough to find, I don't see it happening.

Oh well. Time for me to preorder one of those nVidia 6800 GTs. AGP 8x... yummie.

 

[Insomniak]

Originally posted by: jasonja
As for the jackass ringing the retard bell. I'm well aware that AGP isn't a limiting factor, but in my example there's clearly a theoreticaldifference and bottleneck.

Yes, which means your example is useless (theoretically).

 

[Acanthus]

The sad part is both parts from both companies will perform within 1% of each other, regardless of the solution used. The cards dont need the bandwidth yet, period.