ATI spreading FUD ?

[Kombatant]

Ok just a language thing maybe, but the source of most (99%?) of the data being transferred for the bridge chip will be the gpu.

That's what I meant as well. The VPU does its thing and then sends out the data to the M/B via the PCI Express interface. No matter how fast PCI Express is though, since it's AGP you are limited by that. That's what I thought.

 

[epicstruggle]

Ati's bridge chip <> NVidias bridge chip. They need Nvidias bridge chip and cards to have first hand knowledge.

I am not an engineer, but it seems like there is not very many ways to impliment a bridge chip. Ill defer to those with more knowledge to correct me.
later,
epic

 

[Bjorn]

Ok just a language thing maybe, but the source of most (99%?) of the data being transferred for the bridge chip will be the gpu.

I thought that most of the data (99%?) was being transfered to the GPU from the motherboard. The big thing with PCI Express is that you can transfer data from the GPU to the motherboard in a faster rate. Which might be good for f.e HDTV encoding and also for future games.

 

[AlphaWolf]

Ok just a language thing maybe, but the source of most (99%?) of the data being transferred for the bridge chip will be the gpu.

I thought that most of the data (99%?) was being transfered to the GPU from the motherboard. The big thing with PCI Express is that you can transfer data from the GPU to the motherboard in a faster rate. Which might be good for f.e HDTV encoding and also for future games.

Well actually the advantage of pci-express is that you can transfer data to the gpu at the same time as it transfers data to the motherboard. (4gb/s, both ways). AGP cannot do this, perhaps the bridge chip will make this possible somehow? I dunno.

 

[Bjorn]

Well actually the advantage of pci-express is that you can transfer data to the gpu at the same time as it transfers data to the motherboard. (4gb/s, both ways). AGP cannot do this, perhaps the bridge chip will make this possible somehow? I dunno.

Yep, it's full duplex afaik. And i doubt that the bridge chip will allow this. Though you would have a hard time achieving those figures, mostly because of system RAM limitations.

 

[kth-dp03]

More "in between" circuits = larger latency - which is pretty obvious. Proper implementation will always be faster.

But the question is wheter or not it actually matters. AGP 4x and 8x has already "proven" that the bus is faster than the amount of data which is being sent; since graphic cards these days have so much onboard memory that they seldom have to store data in the system memory.

 

[Bjorn]

But the question is wheter or not it actually matters. AGP 4x and 8x has already "proven" that the bus is faster than the amount of data which is being sent; since graphic cards these days have so much onboard memory that they seldom have to store data in the system memory.

I don't think it matters in the lifespan of NVidias bridged solutions, at least not for gaming. Especially since AGP is going to be on the market for quite some time to come. The NV40 and R420 are native AGP cards f.e, although PCI Express versions will follow.

But the full duplex part is going to be interesting, even for games:

http://www.beyond3d.com/forum/viewtopic.php?t=9986

 

[Evildeus]

Well it seems to be FUD. But well, does it really matter? I mean do you expect more than 1% differencies, performance wise, between AGP and PCI-X cards? I don't.

 

[incurable]

Other graphics companies have cards that are compatible with PCI Express, but they are still only AGP cards that are ?bridged? by a second chip to be physically compatible with PCI Express slots on the motherboard. This architecture can only work at AGP speeds,

AGP is the slower of the two bridged busses, it's the bottleneck. True statement. (note: They don't even say at which AGP speed it'll work)

and is more vulnerable to failure,

More devices mean more points of failure. True statement.

performance bottlenecks

They're always bottlenecks and nVidia's implementation has worse than a native PCI-E 16x interface. True statement.

and incompatibility with software applications.

Oh well, it's software, you'll always find something that wont work. However, this IMO is a stretch.

I don't think ATi's statement is FUD as much as it is marketing's way to spin a technical difference into a perceived advantage in the marketplace.

cu

incurable

 

[Bjorn]

AGP is the slower of the two bridged busses, it's the bottleneck. True statement. (note: They don't even say at which AGP speed it'll work)

There are no official AGP 16X speed afaik. Saying that it's not as fast as a native solution would have been ok.

More devices mean more points of failure. True statement.

More points of failure doesn't automatically mean that it's more likely to fail.

 

[DemoCoder]

The peak speed difference isn't FUD, but IMHO, talk about potential failures and software compatibility is. I could just as easily point out that AGP is a well understood interface, and there is *more* potential for buggy implementations of PCI-E as it is new and immature, this applies to both bridge chips and native implementations.

I expect the next FUD paper to be germinating already, on why FP32 isn't needed, why FP filtering/blending is unusable at 16-bits, how SM3.0 provides no real benefits (atleast for a generation), etc.

 

[DarN]

I expect the next FUD paper to be germinating already, on why FP32 isn't needed, why FP filtering/blending is unusable at 16-bits, how SM3.0 provides no real benefits (atleast for a generation), etc.

That is probably correct. You should also expect papers about how FP32 is absolutely crucial and that SM3.0 is needed right this instant coming from someone else...

This statement is a PR/Marketing reflex-reply to nVIDIA's bridge chip announcement methinks.

 

[epicstruggle]

I expect the next FUD paper to be germinating already, on why FP32 isn't needed, why FP filtering/blending is unusable at 16-bits, how SM3.0 provides no real benefits (atleast for a generation), etc.

I might be naive, but if the above is true, why not make a case for it. If untrue, then why not call it bs?
later,
epic

 

[Kombatant]

I expect the next FUD paper to be germinating already, on why FP32 isn't needed, why FP filtering/blending is unusable at 16-bits, how SM3.0 provides no real benefits (atleast for a generation), etc.

I might be naive, but if the above is true, why not make a case for it. If untrue, then why not call it bs?
later,
epic

I agree. I believe we tend to use the acronym "FUD" far too easily. I mean, FUD means having no proof. If ATI (and in ATI's place could be nVidia) can prove that it has a point, why categorise it as FUD?

 

[Bjorn]

I agree. I believe we tend to use the acronym "FUD" far too easily. I mean, FUD means having no proof. If ATI (and in ATI's place could be nVidia) can prove that it has a point, why categorise it as FUD?

That's true but "marketing departments" and "having proof of something" are two sentences that usually doesn't combine that well.

 

[Kombatant]

I agree. I believe we tend to use the acronym "FUD" far too easily. I mean, FUD means having no proof. If ATI (and in ATI's place could be nVidia) can prove that it has a point, why categorise it as FUD?

That's true but "marketing departments" and "having proof of something" are two sentences that usually doesn't combine that well.

Tell me about it... I love marketing folk, especially those who tend to memorise certain facts to impress ppl, and when you start asking them more specific questions they give you the look

 

[Colourless]

FUD need not be false information. FUD of course stands for Fear Uncertainty Doubt.

I would say that ATI is very much spreading FUD. They want people to Fear bridged PCI-Express solutions, be Uncertain about bridged PCI-Express solutions and of course Doubt the bridged PCI-Express solutions.

 

[Atomahawk]

Having worked in marketing and sales for 10years for a fairly large firm in a complex and competitive environment, I can assure you that no smart company allows the PR or marketing departments free reign as to what can be said to the public, especially when it involves technical or safety issue’s. It just makes no sense too, as most sales and marketing people are not engineers and or chemists.

In my case the chemists explain to us the process’s involved and the benefits to the consumer, we in turn to the best of ability try to translate those benefits into words that MR and MRS can understand. Before anything is actually hard copied out to various media it is looked over by those same R&D departments to make certain that the information is accurate.

Oh and it’s a good thing it’s not the R & D dept. doing the marketing as they would have 10 pages of technical jargon that would almost tell you that if you buy this product you’ll live longer, sleep better, clean everything in your house with a drop etc. Balance is achieved by having separate dept when one markets a product to consumers.

On the same ATI page:

Multiple Benefits of ATI’s True PCI Express Solution

ATI’s PCI Express design provides up to double the bandwidth of bridged PCI Express solutions. Full bandwidth is available in both upstream and downstream directions, whereas bridged PCI Express (AGP) provides only unidirectional bandwidth.
 Better reliability
There are fewer failure points with native one-chip ATI PCI Express due to the smaller number of physical connections, which lowers the time delay between when data is requested and when it is delivered. This also translates into more robust error correction and recovery than bridged PCI Express.
 Better power management
The serial bus with the reduced pin structure of the ATI native PCI Express architecture reduces the number of signals required, supporting lower power consumption and PCI Express's low-power idle states.
Notebook users will find this feature of particular importance.
 More cost efficient
Unlike the bridged chip, the native one-chip design of new ATI PCI Express graphics processors will be brought on stream without significant incremental system cost.

We have to assume that what ATI is saying is in fact true as to how they perceive the benefit to the end user and we cannot automatically regard it as fud, unless you know their product as well as they do, and can prove that this information is in fact incorrect or an exaggeration. Which of course can be done when, we have a working product that you could actually test to see if the statements are in fact accurate. I see nothing wrong with the statement and will take it at face value.

 

[Dave Baumann]

How do you know ati didnt also research a bridge chip and are making assumptions based on their findings? Maybe thats why they can make certain claims, from first hand knowledge. Just my 2 cents.
epic

At the Technology Days thats the first thing they said when they talked about PCI-E bridging - "We looked at a bridge solution and this is what we found..."

As for the "Software incompatibilities" part - it won't have any problems with todays software, but what about software written for PCI-E expecting full downstream and upstream bandwidth?

 

[Evildeus]

At the Technology Days thats the first thing they said when they talked about PCI-E bridging - "We looked at a bridge solution and this is what we found..."

So does it mean they look at all possible bridge solutions? Even on Nv cards?

As for the "Software incompatibilities" part - it won't have any problems with todays software, but what about software written for PCI-E expecting full downstream and upstream bandwidth?

I suppose it could induce issues, but that depends on how the Nv bridge works, how does PCI-X software affect Nv's bridge, does PCI-X software will take into account NV bridge?etc.

So till we know, isn't it FUD as definied by Colourless?