'Fusion' cards

[Rob M.]

superfly: dunno how I missed your post.. let's continue.

<font face="Verdana, Arial, Helvetica" size="2">Originally posted by superfly:
Compiling a game and playing it,are two very different things,and you know it.
</font>

yeah, just like I knew defrag wasn't a game. I was using it to make the point that different programs stress different parts of the system. I still maintain that the current bottleneck for 3d-gaming is the video memory bandwidth (man I'm getting tired of typing that) because, while you can configure quake3 so it's no longer the bottleneck, no one actually plays it like that.

<font face="Verdana, Arial, Helvetica" size="2">
The example i gave still applies,because while getting 100+fps in today's games is indeed overkill,the fact remains that we have cpu's that are already capable enough to play next generation games with 60/70/80 polys per frame,but not the bus to match.at least not until most people have systems wich no longer rely on standard sdram.
</font>

Tell me, what do you think would look nicer, a game running at 640x480x16 with 80k polys per frame, or a game running at the same framerate at 1024x768x32 with FSAA but with only 20k polys per frame? Choose your bottleneck.

<font face="Verdana, Arial, Helvetica" size="2">
Do you really think that of those 500 megs,developers will use it all up in vertex data???,there are other considerations like texture uploads,sound,AI,physics,gameplay,collision,clippi ng,etc..

All those nibble away at those 500 megs of bandwith which can't be completely dedicated to feed just the video card.

So real terms,you won't have much more than those 266 megs available for the video card anyways.</font>

You're right, but, as the p4 benchmarks show, current systems aren't using up all the available bandwidth either.

It seems to me that you're trying to say that when developers start making games designed for systems that have p4 bandwidth or higher, pc133 systems will be bandwidth limited. Well, gee, thanks for pointing that one out, I sure hope that I've upgraded a little by the time p4s are the minimum requirement for games.

I still don't see how you can insist that we're currently bandwith limited.

[Rob M.]

Finally found it, the post where I got my original info about the 16 xeon box id uses, heh they nicknamed it 'fat bastard'.

[superfly]

The highest end video card available today,can already acheive roughly 60 fps at 1600*1200 32 bit in Q3,at that resolution AA is all but pointless.

In fact,even games that are better suited for AA,like fight sims or racing games look incredible at that resolution as well(1600*1200 32bit),and don't benefit very greatly with AA enabled.

A lot of people can play at that resolution not only because of the available video cards on the market,but also because of the price drops that make 19 inch(or better)monitors relatively affordable.

I ask you this,judjing by the progress we've seen so far in video cards,what will happen when within the next 12 months,the available fill rate on video cards will at least double(possibly more),relative to the current ones and every single video card on the market,regarless of who makes it,can play any game at that resolution at 100+ fps.

By then we'll have all the fill rate we could ever possibly want,but even by then,there will still be a lot of users that still will use systems with standard sdram.

What happends then???.....

[Wombat]

Wayne,
It's not inefficient to do larger than 4 CPU's. You're just limiting your thoughts to Intel's performance. If the CPU's and OS are designed with massively parallel operation in mind, things can go very well. It's partly a function of how many pins & traces you're willing to devote to cache snooping.

[Rob M.]

<font face="Verdana, Arial, Helvetica" size="2">Originally posted by superfly:
The highest end video card available today,can already acheive roughly 60 fps at 1600*1200 32 bit in Q3,at that resolution AA is all but pointless.
</font>

60 fps < 155 fps. videocard bottleneck > bus bottleneck.

I don't agree with your 'FSAA not needed' comment though.

<font face="Verdana, Arial, Helvetica" size="2">
I ask you this,judjing by the progress we've seen so far in video cards,what will happen when within the next 12 months,the available fill rate on video cards will at least double(possibly more),relative to the current ones and every single video card on the market,regarless of who makes it,can play any game at that resolution at 100+ fps.
</font>

In the future we will have other bottlenecks, certainly. I definately won't argue with that.

<font face="Verdana, Arial, Helvetica" size="2">
By then we'll have all the fill rate we could ever possibly want,but even by then,there will still be a lot of users that still will use systems with standard sdram.

What happends then???.....
</font>

That sounds a lot like the bill gates' quote "Computers will never need more than 640kb ram". By the time next-gen video cards are standard in retail system packages, DDR will be the slowest memory used in them...

[superfly]

Wrong...suppose you've just bought a ghz p3 or athlon and they still use sdram,you mean to tell me that in less than 12 months,the user will be making a significant upgrade by changing their cpu,motherboard and memory again???

for 90+% of users,not a chance...

You see,as fast as cpu's have been increasing in terms of performance,that's nothing compared to the improvement in video card performance we've seen in just a little over 3 years now where the fastest video cards today are at least 30 times more powerfull that the voodoo 1(in terms of both actual fill rate and added features.)


If graphics chip makers continue the same trend,we'll have a very serious problem on our hands in 12 months time when every single card available by then will be at least 5x more powerfull than anything we have today.

[Rob M.]

Read what I said, there's a HUGE time difference between a computer using a video card as a standard component and it's regular release. The average 3d accelerator shipping with today's computers isn't even in the tnt2 speed range. It'll be at least a couple of years before next-gen cards are std equipment in retail computers, and by that time, sdram will probably be more expensive than both ddr and rdram.

Sure a user can go out and upgrade to a next-gen video card in a few months (which probably still won't overly stress a pc100 machine's system bus), but they could go out and buy a DDR board today.

[Rob M.]

<font face="Verdana, Arial, Helvetica" size="2">Originally posted by frankymail:
I'm positive the own at least one fab (but I think they really own four, but i'm not sure about this)</font>

hrm

<font face="Verdana, Arial, Helvetica" size="2">Fabless Manufacturing

As IC manufacturing processes become ever more complex and capital-intensive, VIA's close partnerships with the world's leading silicon foundries and packaging and testing houses are assuming increasingly vital importance. These partnerships not only allow for the most up-to-date production and testing technologies while defraying the immense costs of purchasing this equipment, but also ensure guaranteed production for VIA's products while providing the flexibility its customers need to ramp up volumes to meet sudden changes in market demand.

VIA's own Manufacturing Engineering and QC Teams work closely with its manufacturing and assembly partners to ensure world class levels of product quality and reliability. All chipsets, processors, and communications chips shipped by VIA undergo the most exact testing procedures possible, and VIA's commitment to quality is demonstrated by its ISO 9001 certification as well as its meeting the rigorous supplier requirements of leading PC OEM customers.

VIA's close connections with its foundry, assembly, and testing partners are the key to the success of the company's unique fabless business model, enabling it to focus on maximizing architectural partitioning, unit performance, volume/price ratios, product quality and reliability, and volume production. This means that VIA delivers products of the highest quality with the latest technology in the most cost-effective manner. </font>

In the future, please don't say you're sure about something, unless you really are sure. It makes it very hard to believe anything you're saying is correct..

[Ali]

RobM, I wouldnt trust that mention of fat bastard as far as I could spit it.

Its from some freelance Flash animator who prob heard a rumor from someone about a 16 way processor, and assumed it used xeons.

Not that I realy care, but after spending a whole 5 minutes looking up the information on that 16 way processor, I didnt find one mention of a new 16 way computer.

Ali

[frankymail]

Hi,
Hey Rob, you were right, yet you were wrong; Via itself has no fabs, yet they do: they have bought Cyrix a long time ago, and Cyrix had fabs. Via is currently in the process of selling those fabs and contract UMC and TSMC to fab their chip... yet TSMC might be temporarily crossed off the list, because they have to much patent litigation going on... They plan being fabless by the end of 2001; being fabless will allow them to get the lowest fabs contracts and to sell the CPU for the lowest price point possible... http://www.eoenabled.com/edtn/out.as...s%2C+core+logi c ; I know they are on the verge of becoming fabless, but Via is not stupid enough to waste valuable and expensive equipement, and also-as-valuable production time...

I know that the company contracted to fab Matrox' chip is NEC; I wonder if NEC is equipped for 0.13/0.15µ designs . . . ? ? ? If they do, then Matrox has no choice other than released the CTWOTA (pronouced See-Tea-Woa-Ta), which stand for "Chip That Will Own Them All"

I hope Big M (Matrox, not McDo) will soon end this tread by amazing us all with a nice new all-mighty board...

Francis Beausejour

------------------
What was necessary was done yesterday;
We're currently working on the impossible;
For miracles, we ask for a 24 hours notice ...

[Rob M.]

<font face="Verdana, Arial, Helvetica" size="2">Originally posted by Ali:
RobM, I wouldnt trust that mention of fat bastard as far as I could spit it.

Its from some freelance Flash animator who prob heard a rumor from someone about a 16 way processor, and assumed it used xeons.

Not that I realy care, but after spending a whole 5 minutes looking up the information on that 16 way processor, I didnt find one mention of a new 16 way computer.

Ali</font>

Actually, TTimo is responsible for maintaining the public release of q3radiant, the quake3 level editor that everyone uses. He also helped id's Robert Duffy create it. I think we can trust he's not lying in this case.

[Jammrock]

Come on! We're almost at MURC's first 500 post thread!

Jammrock

[Rags]

Ummm, Jammrock, MURC has had a couple of 1000+ post threads

Rags

[ccoltas]

March is just over 2 weeks away!

Hope there's a new Kick-ass Matrox card coming out

[frankymail]

"Father I must confess, for I have sinned..."
I'll buy (or at least pre-order) a Millennium G800 64 MB as soon as possible, but in the mean time, I bought myself a Radeon LE 32 MB DDR (128-bits) which is available for $140 CAN!!!(The Radeon LE card is exactly the same as the regular Radeon 32 MB DDR, except it doesn't come with a fan and the chip is only cooled by an heatsink, which I immediatly replaced. I comes with the same 6 ns memory, but has HyperZ disabled in the registry (use the reg hack...) and is clocked down to 150 MHZ core / 150 MHz memory). I can say that I am very happy with my purchase (I overclocked the card to 200 MHz core / 200 MHz memory...) which makes me feel even more shameful... (I've always considered myself, and still do, as a hardcore MURCer... Of course, I kept my G400 in my main box, (I have a Sony E400 and a Sony 210GS attached to it), but I needed a little more for the time being. But, the G400 IS STILL MY MAIN CARD IN MY UNIVERSITY BOX (the Radeon is in my week-end fun box at home), and if I were to choose between the two, I would go for the G400 for Dual-Head and 2D quality. However, I just felt you guys should know that if you're looking for a temporary or second solution that has 3D speed neqar that of the GeForce 2 GTS, as reasonnable 2D/3D quality (much better than the GeForce 2 MX/GTS, yet nowhere near the output of the G400) and is cheaper than the GeForce 2 MX (about $90 US or $140 Can...), this card is worth a look... Go ahead flame me . . . but at least I never even had the thought to go out and buy a GeForce 2 MX/GTS or even Ultra...

Francis Beausejour

------------------
What was necessary was done yesterday;
We're currently working on the impossible;
For miracles, we ask for a 24 hours notice ...

[This message has been edited by frankymail (edited 19 February 2001).]

[This message has been edited by frankymail (edited 16 February 2001).]

[This message has been edited by frankymail (edited 16 February 2001).]