Isn't Nvidia going the wrong way? (No, I'm not bashing Nvidia)

[NFS4]

And if the Rampage were to use two 30 million transistor chips (and both NV20 and Rampage use some form of HSR) the difference would be...:Q

nVidia is setting themselves up for the OEM market by sticking with a single chip architecture.

As for what they have in store to combat Gigapixel technology, we'll have to wait and see before speculating a year in advance.

 

[lsd]

You clearly don't understand how marketing works. Step away from the technical view and take a look at the economical view. What your saying doesn't add up right?

 

[SilverBack]

What an interesting thread.

1. Companies aren't made profitable by product alone, they are made by customer perception.
2. Companies are also made by having "the" only product available causing demand.

These are very simple and basic rules that I'm sure that all of you understand.

How does the largest sector or "pool" of video card buyers thinks about each product?
What are the advantages to either and what do they have in common?

DaveB3D seems to think that NVidia is coming up to a theoretical roadblock for thier current implementation of video cards. Didn't 3DFx just prove that they had hit thiers? ( No V5 6000 )

Tile rendering isn't new, it's just something that is technologically very hard time to implement.

In the next few months Nvidia will have to change, but they haven't come to any unsurrmountable impass. They will move to a .13 micron die manufactoring process in the coming year. They will more than likely also change to QDR ram which is currently under development. ( QDR , quad data rate ram )

The arguments made in favor of 3DFx are getting old. They can't be defended well either.
The company failed to realize the importance of thier own customers. Read number 1 of above.
For months 3DFx preached that 32 bit color wasn't needed. The customers demanded it, Nvidia delivered. Customers are now demanding T&L, will 3DFx have this on thier next card?

This really isn't thread on technology, it's a thread about customer loyalties.

BTW the 100 fps at 32bit color at 1600x1200 for the NV20 is probably realistic. The NV20 will have a 256 bit data path, if it does come with QDR memory it has more than enough memory bandwith to easily do that.
Also my GTS Ultra already gets 61 fps in the same situation, so 100 fps isn't unreasonable at all.

 

[ndee]

BTW I thought that the Q3A Benchmark was just a speculation of this www.geocities.com/XYZ, wasn't it?

 

[BenSkywalker]

This doesn't sound like Intel vs AMD, this is an exact replay of PPC vs x86.

"NVIDIA's methodology is flawed"

Replace all instances of nVidia with Intel in this thread and it is an exact replay of the PPC/X86 debate that was so hot several years ago. Now we have the two companies with the "flawed" methodology dominating and massive layoffs from the company on the "right track", along with a roughly 95%/5% split of the market respectively.

Engineering theory and actual real world marketplace performance are two very different things. The PPC design IS a clearly superior design looking at it from an engineering aspect, engineers are often too wrapped up in the finer points to see the bigger picture. Not to mention, the real world performance of PPC lags behind x86 significantly, despite all the "obvious" advantages it has, and the fabrication "brick wall" they were supposed to run into several years ago(not making this up, if anyone cares to look you may be able to find some old articles stating as much, around '95-'97 it was a hot topic).

The "brick wall" thoughts require that Moore's law cease to be a ~reality. Sooner or later we may well hit that point, but next year? I would bet against it.

 

[NFS4]

<<
Engineering theory and actual real
worldmarket place performance are
two very different things. >>


And that's been my whole point when arguing with Dave. He keeps looking at the technological side instead of the economics, market execution, and the &quot;big&quot; picture. Like I've said before nVidia isn't the one with something to proove right now, and it takes more than one product to turn a company around.

STOP THINKING LIKE AN ENGINEER FOR ONR MINUTE!

 

[fodd3r]

nvidia's design is lame. it's getting transistor happy and they're chips are getting far too complex. it's pretty clear that nvidia is really a bunch of engineers recruited from sgi --nvidia stole a fair bit of talent from them, stupid sgi, they're just like micron and xerox, amazing technology and they don't execute-- look at how massively parallel each of their chips are, 4 pipelines!

a few impacts of the transistor happy nature.
-more heat, which means added cooling cost.
-more power means greater work needs to be done to shield from emi and other such things. again, more cost.
-greater production cost, each new chip design inccurs an overhead.

as it stands sgi, erm... i mean nvidia is pushing out some nice chips. but their transistor happy designs are starting to catch up with them, the geforce 256 was evidence.

before we start bashing multichip solutions get the facts straight. if 3dfx was in god's good graces they'd have excuted on time. their vsa-100 would have trounced over most anyone. here is why:

it's all becuase they have a single scale able architecture.
-one chip means lower production costs and production investments. you don't have to build a new chip for every market. this incurs a massive overhead reduction --good for margins.
-a simple chip means high yields. --too bad 3dfx didn't get their design process right, if only they went .22 micron --s3 did this, i think-- it would likely have been enough. *mutter*
-oems like this, since they can have one chip that handles both high end and low end. with little added cost! this is with pcb considerations taken into account. you may say no way to this, but if you consider it this way, you could buy a helluva lot of chips in bulk, then as demand calls for it use them in single or multichip solutions. what does this mean, if you ordered a lot of one chip and it turns out demand sucks, well then your screwed you've burned lots of money, are wasting warehouse space and having idle production lines --this last one is dependent on the oem.

as to why it didn't work out in 3dfx's favour, well their design methodolody and delayed excution screwed them good. otherwise their vsa-100 is rock solid.

 

[BenSkywalker]

&quot;nvidia's design is lame. it's getting transistor happy and they're chips are getting far too complex. it's pretty clear that nvidia is really a bunch of engineers recruited from sgi --nvidia stole a fair bit of talent from them, stupid sgi, they're just like micron and xerox, amazing technology and they don't execute-- look at how massively parallel each of their chips are, 4 pipelines!&quot;

nVidia is handling quite a bit of SGI's R&amp;D as of late, not to mention that you would be hard pressed to find any 3D hardware company that doesn't have any former SGI employees.

&quot;a few impacts of the transistor happy nature.
-more heat, which means added cooling cost.&quot;

Just like Intel and AMD

&quot;-more power means greater work needs to be done to shield from emi and other such things. again, more cost.&quot;

Just like Intel and AMD

&quot;-greater production cost, each new chip design inccurs an overhead.&quot;

Just like Intel and AMD

So far, you are making a very strong case for nVidia, what exactly is supposed to be bad again?

&quot;before we start bashing multichip solutions get the facts straight. if 3dfx was in god's good graces they'd have excuted on time. their vsa-100 would have trounced over most anyone. here is why:

it's all becuase they have a single scale able architecture.&quot;

Hmm, sounds an awful lot like PowerPC

&quot;-one chip means lower production costs and production investments. you don't have to build a new chip for every market. this incurs a massive overhead reduction --good for margins.&quot;

Just like PowerPC

&quot;-a simple chip means high yields. --too bad 3dfx didn't get their design process right, if only they went .22 micron --s3 did this, i think-- it would likely have been enough. *mutter*&quot;

Just like they thought PowerPC and PPC would, too bad it never worked

&quot;-oems like this, since they can have one chip that handles both high end and low end. with little added cost!&quot;

Who? Apple? I'm assuming that all of the &quot;farming&quot; OEM work other companies will do building 3dfx parts will maintain the standard sell through sales payment technique. This protects them from supply/demand/price fluctuations. It is fairly standard and is used in pretty much everything now.

Your arguments above are a great summation of everything the PPC and PowerPC advocates said year after year. IBM's and Motorola's microprocessor revenue combined likely doesn't equal that of the weaker x86 company, let alone Chimp and Chipzilla combined.

You are using thoughts of engineering principles and trying to apply them to the marketplace, things don't work like that. If they did, we sure as he!! wouldn't be running x86 processors on mobos still having ISA slots running a Win9X based OS(well, that describes the majority of people anyway).

 

[Blackhawk2]

BenSkywalker, Nvidia is in the process of producing quantity over quality. Everyone knows 3dfx's 2D/texture quality is superior to Nvidia's. Nvidia's Geforce 256 has a hardware flaw that prevented its T&amp;L unit from working at top perfromance, this was corrected in the Geforce2. Geforce 256 also had instructions to slow it down because it was overheating (note: this may happen with NV20->15 watts is very close to the same wattage as the Geforce 256).

Nvidia is clearly on the road to heavy transistor counts, high heat, high power consumption and all the problems/costs associated with this type of roadmap.

Notice how AMD has taken alot of flak over the amount of heat being produced by Thunderbird/Duron and is now trying to move to a more pure silicon process and lower micron size to significantly reduce the heat dissipation by Palomino, Morgan ect.

Also note how one Nvidia board manufacturer was saying that NV20 might require a special cooling system, that is a sign of bad design.

 

[OneOfTheseDays]

Seriously Dave, i think that you are thinking WAY TOO MUCH INTO THE FUTURE. I understand your arguments about the nv20 and they are valid, but we don't KNOW what Nvidia is up to for its later chips. Its way too far off to speculate.

 

[OneOfTheseDays]

Please, the Voodoo 5500 was a SERIOUS design flaw. Any card that needs its own power supply is just a joke. It produces a crapload of excess heat too. It seems that 3dfx went the brute force way by trying to stack as many vsa chips into one board as they could.

 

[Blackhawk2]

Well Sudheer Anne the most problems I see being posted are problems with Nvidia cards crashing/overheating/conflicting with mobo's ect. Theres the quantity over quality argument rearing its head.

 

[Wingznut]

What's up with people buying into a graph that some schmuck posts on a GEOCITIES website?

Besides, didn't you guys notice that (according to that graph) the NV20 is only ~7fps faster than a V5-5500 at 10x7x2xFSAA? And it's ~10fps slower than a V5-6000. Doesn't make any sense.

 

[DaveB3D]

sigh... I wish I could do a mind connection of some kind with you guys.. would be so much easier and you'd be sooo much clearer. the best I can do though is promise you that I know with a fact I'm right (if I didn't, I never would have come to work at 3dfx... as I knew were both companies where headed well enough before i came) and you'll see what I mean in the future...

 

[BenSkywalker]

Blackhawk2-

&quot;BenSkywalker, Nvidia is in the process of producing quantity over quality. Everyone knows 3dfx's 2D/texture quality is superior to Nvidia's.&quot;

Huh? The only thing I can think of that the V5 has an edge in as far as texture quality goes is Quake3 with TC on. To counter that, the GF trounces the V5 in UT and Evolva(the Radeon does em all great, so I would give that round to ATi). 2D quality it isn't all GF based boards, and it isn't all monitors. That isn't an issue with the GeForce, it is board level.

&quot;Nvidia's Geforce 256 has a hardware flaw that prevented its T&amp;L unit from working at top perfromance&quot;

Complete and utter bullsh!t. I bought the GeForce for its' hardware T&amp;L above all else, I work with 3D visualization. The GeForce hits very close to its' theoretical peak given the proper circumstances, closer then any other board based on what I have seen(the Radeon blows in this comparison).

&quot;Geforce 256 also had instructions to slow it down because it was overheating&quot;

Which the only time I have ever seen kick in is when a board is OCed too far. Looking at OCing, the GF2 boards are much better then the V5, I don't think anyone will argue that point.

&quot;Nvidia is clearly on the road to heavy transistor counts, high heat, high power consumption and all the problems/costs associated with this type of roadmap.&quot;

Hmmm, who has the highest heat generation and highest power consumption right now? 3dfx. Who just launhced the fastest notebook 3D accelerator ever, which uses an amazingly low amount of power? nVidia. Who is headed where?

&quot;Notice how AMD has taken alot of flak over the amount of heat being produced by Thunderbird/Duron and is now trying to move to a more pure silicon process and lower micron size to significantly reduce the heat dissipation by Palomino, Morgan ect.&quot;

And I run an Athlon and am upgrading too a T-Bird. I couldn't care less about how much heat it generates. Does it work? He!! yeah. Is it fast in 3D? You bethca. Is it stable? Yep. As a consumer, why should I care about anything dealing with heat or power consumption as long as it works?

&quot;Also note how one Nvidia board manufacturer was saying that NV20 might require a special cooling system, that is a sign of bad design.&quot;

Bad design?!?!?!? Is that a joke? Do you also consider the GHZ PIIIs and Athlons bad designs? They also needed special cooling. Who cares? They work(the T-Bird/Athlon always did too, even though they were the hotter/more power hungry of the two).

&quot;Well Sudheer Anne the most problems I see being posted are problems with Nvidia cards crashing/overheating/conflicting with mobo's ect. Theres the quantity over quality argument rearing its head.&quot;

Crashing/overheating/conflicting? Could you please post some links. Problems with mobos? How about the V5's inability to fit in the PIV motherboards, that is a rather serious issue. Do I think it is 3dfx's fault? Absolutely not. Every properly built motherboard works just fine with all the cards, as they should.

Dave-

I'm sure you believe that to be the truth. The fact is that you are working for a company that has earned a very strong reputation for failing to execute, and are trying to convince everyone that they are going to &quot;take down&quot; the most successful video chip manufacturer ever. The reasoning behind this seems to be that they are moving to a more &quot;elegant&quot; design philosophy. That hasn't worked yet in the PC industry. There is a big difference between having the best designed by engineer's standards products to having the widest marketplace success. Given 3dfx's defense of not having AGP texturing with the V4/V5, it seems clear that they still don't grasp what it will take to truly take control of the market.

 

[DaveB3D]

AGP texturing?? What?

That has not proven itself yet by a long shot... and that still does not prove anything. The only proof I can offer is my complete and total confidence, which i hope is worth something. If I wasn't completely confident, I never would have come here in the first place.

 

[Sunner]

Lets look at the current situation.
nVidia has gone from a virtual noone to being king of the hill, and has been making extremely strong results for like 2 years in a row now, and have even managed to overtake ATi as the OEM king.
And they are expanding into other areas such as chipsets, and sound.

3dfx OTOH have gone from being undisputed kings of 3D to being the noone, with near zero OEM precense(sp?).

ATi has lost alot to nVidia, but have gained alot in the Retail market OTOH, and they're still strong in the OEM sector, and they are, like nVidia, expanding into other markets.

As for brute force vs more stylish ways, who cares?
Like Ben said, if it works, who cares how it works really?
Ok, it can be interesting to talk about the inner workings of chip, so in that way one could care, but when playing UT or Q3, all you really care about is looks and speed, right, now the amount of transistors.

One might argue that the GTS is a brute force only aproach, but lets face it, it works, and its on the same price level as the Radeon and V5, so who cares?

And like someone said, considdering their current position, 3dfx needs to prove themselves, not nVidia, nVidia has proven themselves for 4 generations in a row now, that ough to do it.

 

[SilverBack]

He's right, current 3Dfx cards only support the AGP bus speeds, AGP texturing isn't used at all.

Dave, this is my opinion and mine only, this is possibly the worst time for anyone to even remotely be interested in gaining employment with 3Dfx.
With consecutive bad quarters ( hell, years now , 2 ) 3Dfx is most likely on the thinest ice it has ever been with respect to thier business prospects.
ATI is the leader with Nvidia right on thier heels.

It doesn't matter at all who has the best technology, it matters who can produce the results to stay active in the market. It won't matter how good 3Dfx's next generation product will be if it never makes it to the retailers. 3Dfx needs a new product out now.
We all kept hearing all summer that 3DFx had the number one selling video card, but they couldn't turn a profit and stock prices fell. Thier chipset however was no where near the chipset sales that Nvidia gained.

Unless 3Dfx has a new card out by the first quarter of 2001 that is compelling to buy, you might as well right them off. It's to bad really to. We would all like to see them come back to the original standards that made them into the 3D power they once were. Poor leadership and just dumb mistakes have put them in this position. The only thing that will put them back on track is a card that dominates the rest of it's competitors. Frankly many of us don't see that happening anytime soon.

 

[BenSkywalker]

&quot;AGP texturing?? What?

That has not proven itself yet by a long shot... and that still does not prove anything. The only proof I can offer is my complete and total confidence, which i hope is worth something. If I wasn't completely confident, I never would have come here in the first place.&quot;

The fact that my comment on AGP texturing needs any further comment is a good indicator why 3dfx is in the position they are. What kind of return are they seeing on R&amp;D and what useful life are they seeing out of the products they develop?

Long after then V4 and V5 are done with, and 3dfx is producing Rampage and GP tech boards, nVidia will still be making money, a lot of money, off of the GF2MX offerings. Their low end mass quantity parts fuel their R&amp;D which in turn gives them superior offerings to move down into the lower end of the market.

Without any support for AGP texturing, 3dfx has eliminated any reasonable hopes of becoming an integrated solution. OTOH, nVidia, ATi, Matrox and S3 have all left this oppurtunity open. Dominating the high end market is a very small aspect of the marketplace in its' entirety. By taking over the low end solutions, nVidia has the advantage of scales of economy to help fuel them. Same with ATi, though as of late nVidia is the one with the momentum(could change with ArtX).

Having faith in your confidence would require two things. One is that 3dfx can deliver as scheduled. That is something that they have failed to do on a very large scale. A year after the V5 6K was supposed to ship they finally canned it, doesn't inspire confidence.

The other factor is that you are hearing the truth from nVidia. It is very common in the business world to &quot;leak&quot; details that would be benefitial for you to have your competitor believe. nVidia has proven time and time again that they are not a stupid company, and I know d@mn well that I would make sure that certain &quot;facts&quot; were leaked into the proper hands. Things are not cut and dry when dealing with competitors as you may think. You may be hearing from your friends at nVidia exactly what they want you to. I would think hard about the tactics nVidia is known to use before you dismiss this too quickly.

 

[DaveB3D]

na... I know that isn't the case.. Why? For the simple reason that much of what I know comes from pre-3dfx days.

 

[BenSkywalker]

&quot;na... I know that isn't the case.. Why? For the simple reason that much of what I know comes from pre-3dfx days.&quot;

I assumed that to begin with.

 

[DaveB3D]

So they are going to tell web sites false information? Nope... I know this not to be true, btw, because things that I've been told have proven themselves to be true already.

 

[BenSkywalker]

&quot;So they are going to tell web sites false information? Nope... I know this not to be true, btw, because things that I've been told have proven themselves to be true already.&quot;

Big difference between six months out and a year and half. Some sites were told the NV20 was going to launch this fall early in the year, and late last year. That clearly didn't happen. Saying what is completely set in stone is one thing, for instance the specs on the NV20 are very unlikely to change compared to what you have heard. However, NV25 and NV30 could be a very different situation.

 

[DaveB3D]

The difference is that most sites went through PR.. I know people throughout the company.

 

[BenSkywalker]

Perhaps I should put it like this-

If Jeff got info from 3dfx on Gigapixel, no matter who it came from, I wouldn't believe it.

If Kyle got information from nVidia on NV30, no matter who it came from, I wouldn't believe it.