Poll on future dx9 card's ( not ati and nvidia ) performance

[Entropy]

Would you be hardpressed to answer to a question wether today´s high end cards are rather fillrate or bandwidth "hungry", what would you say?

This question comes up over and over here, and I just don't think the question has a general answer. If we refer back to previous products, they have always tried to strike a balance that suited the expected applications. If we change the expected application behaviour, we can't really draw many conclusions from previous products. Does a 9700 generally perform better than a 9500Pro? Yes, but the degree is very dependent on what you ask of the cards. The DOOM 3 benchmarks suggested that the 9500Pro was limited by its relative lack of bandwidth even without AA, but is that game typical for future apps? We just don't know. We can have an idea of general trends, but that's about it. One thing is for sure though, if you aim for the highest possible performance, it makes sense to simply move the boundaries as far as it is economically possible. Having more bandwidth available is always useful. Spending $20 more to remove bottlenecks for some scenarios, and removing worry that your predictions about future application directions might be a bit off is probably not worth it for a budget design obviously, but may well be a good idea for a would be high-end contender.

Entropy

 

[Joe DeFuria]

It wasn't [PowerVR's] responsibility that 4 died an unexpected and quick death.

Not to nit-pick and beat a dead horse, but, well, I will.

IMO, it is PowerVR's responsibility.

It is PowerVR's decision to use the IP licensing model which ultimately enabled the unexpected and quick death. Had powerVR decided to invest in, fab and sell the chips themselves, we might have seen Series 4.

 

[Ailuros]

Under unrealistic conditionals if my grandmother would have testicals, she'd be my grandfather

 

[Joe DeFuria]

Under unrealistic conditionals if my grandmother would have testicals, she'd be my grandfather

Interestingly enough...

If PowerVR had any testicles, they would have a fabless semiconductor model for the PC business. 8)

 

[Ailuros]

Having more bandwidth available is always useful.

More bandwidth yes, but it depends how you reach it. One of the smallest consideration with TBDRs is the lack of bandwidth.

Heck if we're so bandwidth hungry they could also use a 512bit bus and have the crappiest read efficiency you could imagine. So while you end with more bandwidth it'll turn into a boomerang against you and not for the benefit you'd expect.

Exaggerated example, but IHVs do have their internal tests and predictions, wherefrom the results dictate the decisions they make and what they consider to be optimal under most conditions.

 

[Ailuros]

Under unrealistic conditionals if my grandmother would have testicals, she'd be my grandfather

Interestingly enough...

If PowerVR had any testicles, they would have a fabless semiconductor model for the PC business. 8)

I would love to see a couple of million things myself, ImgTec becoming a fabless semiconductor amongst them. In the end it would be my greed to have a high end TBDR every year, versus the possibility of over 200 people staying unemployed.

 

[Joe DeFuria]

In the end it would be my greed to have a high end TBDR every year, versus the possibility of over 200 people staying unemployed.

Right. It's a risk. And again, the fact that that risk is notbeing taken disappoints me from PC enthusiast standpoint. Not to mention that if they would be successful, there's the possibility of gainfully employing thousands of new employees...

 

[parhelia]

i gave my vote for deltaChrome: that part looks to be a very well ballanced dx9 desktop contender

If you say so, because we have no idea of the clockspeed, features, etc. All we know is that it's a 0.13 micron 8-pipeline GFX engine which supports DX9, a bit vague IMO.
We also know XGI 's XG series is 0.13, has 8 pipelines and supports DX9, so it's back to square one.

And Deltachrome is depicted by S3 themselves as mainstram/low end :

We will leave the hardcore minority to ATI and nVidia, we'll focus on the normal folks with normal budgets. http://www.homelanfed.com/index.php?id=12478.

plus its expected availability is somewhere around this christmass,

I remember this :
Tuesday, January 7th, 2003 : S3 Graphics Launches All New DeltaChrome with Dazzling Hi-Def Graphics and Video with Microsoft DirectX 9.0 and HLSL support
http://www.s3graphics.com/pressrel/2003_01_07.html

Nine months later and still nothing...

 

[Ailuros]

That doesn't change the fact that for the time being at least and their financial situation it is not an option. Now you can either accept that or keep pondering on the same tired issue, like a 2-year old crying it's heart out in front of a candy store insisting it wants chocolate, when the store is actually closed.

Being a PC enthusiast myself I don't see myself being stripped of alternatives by the way and I never personally believed that TBDRs can perform miracles either; I just happen to like the approach that's all.

 

[Captain Chickenpants]

Don't forget that PC graphics is only one part of our business. It makes no sense to treat most of our products as IP, and one isolated product differently.
This is why as a company we can afford to be working on so many projects, whilst other companies are working on fewer projects and are ten times the size.




Just my opinion, not necesarily that of IMG.


CC

 

[Ailuros]

DeltaChrome does look like an interesting contender at least on paper. I can't say I was dazzled with XabreII's 4xAA/4xAF limitations to be honest.

But it's all moot (and that goes for all of them) until we see them tested and analyzed by independant sources. Both above even with a higher clock speed than initially intended look more like mainstream parts at best these days, but that doesn't make them less interesting either.

 

[parhelia]

Now you can either accept that or keep pondering on the same tired issue, like a 2-year old crying it's heart out in front of a candy store insisting it wants chocolate, when the store is actually closed.

Hey, I didn't bring Xabre up in this topic at all, tell me where I mentioned the name Xabre in this topic before.
I was only talking about DeltaChrome and it's from S3. And also about PowerVR.
On the other hand you did :

As far as Xabre goes, it can muster almost equivalent average framerates in 3DCenter's Pyramid2003 demo for UT2003

Moreover, I never said Xabre was better than a GF4 or Radeon 8500, we all know it's a low end card and was aimed towards MX cards...

I was dazzled with XabreII's 4xAA/4xAF limitations to be honest.

Where did you get this info?
I have yet to see any official specs from XGI.

 

[Ailuros]

Don't forget that PC graphics is only one part of our business. It makes no sense to treat most of our products as IP, and one isolated product differently.
This is why as a company we can afford to be working on so many projects, whilst other companies are working on fewer projects and are ten times the size.




Just my opinion, not necesarily that of IMG.


CC

Could any of you shed some light what the purpose of OPS is for and how companies are going to benefit from it? Yes I did read their homesite and their rather vague explanation, but it's not really enlightening for the layman.

 

[Joe DeFuria]

Don't forget that PC graphics is only one part of our business.

I'm not forgetting that at all!

It makes no sense to treat most of our products as IP, and one isolated product differently.

That all depends on if you want to compete in the PC space or not. In other words, if you're content to try and just compete at the integrated PC space, and maybe the low-end discrete, then I would agree it doesn't make much sense.

If you want to compete in the high-end, or "top to bottom" in the PC space, then it makes all the sense in the world to me to isolate one product line differently.

This is why as a company we can afford to be working on so many projects...

And what percentage of these are actually productized?

whilst other companies are working on fewer projects and are ten times the size.

It may also be why other companies that are working on fewer projects, actually seem to ship a higher percentage of those products being worked on....

I really do fully appreciate the "pros" of the IP model. I just think the "cons" outweigh the pros specifically in the PC space. I don't begrudge PowerVR for sticking with IP in and of itself....I just think that if you want to compete in the high-end PC space, (honest question: do you?), the IP model is a hindrance.

 

[Ailuros]

Where did you get this info?
I have yet to see any official specs from XGI.

I said XabreII mind you. No idea yet about XGI and whatever they plan to release.

Moreover, I never said Xabre was better than a GF4 or Radeon 8500, we all know it's a low end card and was aimed towards MX cards...

Nope you didn't, but you exquisitely pondered on the lack of a T&L unit on PVR's products, when other products seem to have them either only on paper or they're completely worthless. ATI and NV being of course the exeptions.

 

[parhelia]

Nope you didn't, but you exquisitely pondered on the lack of a T&L unit on PVR's products, when other products seem to have them either only on paper or they're completely worthless. ATI and NV being of course the exeptions.

But that would include S3 then

 

[Joe DeFuria]

I said XabreII mind you. No idea yet about XGI and whatever they plan to release.

IIRC, (somewhere posted in this forum) XGI will "launch" their first product this month, with other products coming later this year. I personally get the feeling that the first XGI product is, in fact, the XabreII.

I don't think we'll get a good feel for XGI's plans until they launch their second batch of products.

 

[parhelia]

Who tells you XGI hasn't changed the specs ?
After all, they didn't announce anything, no specs given yet, which gives them the freedom to change anything they want.
And there may be more to it than a simple renaming.

 

[Ailuros]

I really do fully appreciate the "pros" of the IP model. I just think the "cons" outweigh the pros specifically in the PC space. I don't begrudge PowerVR for sticking with IP in and of itself....I just think that if you want to compete in the high-end PC space, (honest question: do you?), the IP model is a hindrance.

All the indications so far from their publicly announced AGM meetings feedback dictate that their forecasts and targets are to increase the volume of developed chips at the same time from about 10 today to 30-40 (under ideal conditions) in the foreseeable future.

Would PowerVR be a completely independant entity, it might be an option. I'd suspect that even attempting to either part it from the other business core or even sell it wouldn't be that easy anymore. I'd be pretty naive to believe that Ensigma and Metagence hasn't in fact ANY active involvement in whatever IP PowerVR develops and vice versa.

 

[Joe DeFuria]

Who tells you XGI hasn't changed the specs ?
After all, they didn't announce anything, no specs given yet, which gives them the freedom to change anything they want.
And there may be more to it than a simple renaming.

Of course, anything's possible.