Beyond3D Wolves Needed to Tear This One Apart

Status
Not open for further replies.
First thought from a relative outsider to this business. You describe nVidia as introducing technology each step of the way with the GeForce series.

I must admit my impression was somewhat different. It seemed more like every second generation introduced technology which usually had performance issues. The next generation then refined that to produce a good part. You can see this with the TNT which was an okay part but refined into the TNT2 which was an excellent part, GeForce which had T&L but that really took off with the GeForce 2 etc...

Also geometry instancing isn't something that only the newer R420 designs can do. Your article kinda implies this but as I understand it most R300 products can do it quite happily.
 
Evidence? You expect me to have evidence on a speculative piece! :LOL:

Late Spring of 05 was the last hint and rumor that I had heard. Admittedly I would believe that NV50 would be more of a Fall 2005 product, but late Spring was the last whisper that I had heard.
 
How does that reconcile with the public statement that the architectural cycle will be in the order of 18 months?
 
Haha, when a certain engineer we all know scoffed at that and said, "I haven't seen it slow down to 18 months yet."
 
Haha, true. I guess I am also basing this off of another individual I know that works very close to NVIDIA that said, "NVIDIA internaly declared war on ATI, and it is honest-to-God no holds barred combat." I think NV might try to be really agressive this next Spring. Now, whether they can actually get significant quantities of parts out at that time...

Then again the actual reign of the NV3x technology (from when it was first sold on the market as the GeForce FX 5800 to when the NV40 was released) was about 14 months- nevermind that it was initially introduced in November of 2002!
 
JoshMST said:
Haha, when a certain engineer we all know scoffed at that and said, "I haven't seen it slow down to 18 months yet."
DaveBaumann said:
:oops: NV2x lasted in excess of two years!??
:LOL:

Josh, I got to give you some big points for coming here and looking for input/criticism...'specially after I re-read your last piece and remembered who ya were! ;)

It's better, but there is still a bit of a green-tinted bias to it. I'll get more specific in a bit, I gotta go do the dinner thing for the kids.
 
Yet another piercing question! I thought I had mentioned late Holiday Season 2004 in the article. Also, why is NV48 mentioned in the latest leaked drivers, but not NV47?

Digital! Green tinted bias! Haha,I can't wait to see the justification for that one.
 
OK, when whomever you spoke to talked about a Q2 '05 / Spring '05 did they specifically mention NV50 or something along the lines of "their next generation part"?
 
No, they did not specifically mention NV50.

Honestly Dave, I am not trying to misguide or push some kind of agenda, I am just working off the best information I could dig up and glean from anyone I could.
 
ATI claims that the R420 has around 165 million transistors, but it appears as though they are not counting every transistor on that die. By rough measurements, the die size of the R420 is slightly smaller than that of the NV40.
I remember a discussion about what is and isn't normally counted as a transistor, was ATi not counting things or was nVidia counting bits that normally weren't counted? :|

ATI is very busy at building a solid community of users, and they are doing this while still protecting their IP and image. By building such a community there will always be a very loyal consumer base that will support the company even when their products may be second best to whatever else is out there.
Uhm, when have their products been second best to whatever is out there? Since their company started to not suck at about the same time that they started reaching out to the community I'd like to think it's more of a corporate philosophy change rather than a PR move to help them in what you imply is their impending downfall.

NVIDIA also looks to introduce the NV50 in late Spring of 2005. What this will bring to the table is unknown. While sketchy details of the R520 are available, very little is known about the NV50.
rofl.gif
rofl.gif
rofl.gif


There is a reason that there are some details available about the R520 and damn near nothing about the nV50....but I'll leave you to speculate on what that reason may be.

Ok, one hint. Your almost right on the time, but waaaay off on the product. ;)

ATI has not stood still through all of this, and the X800 and X700 series of products match NVIDIA step for step.
"Match nVidia step for step"? Uhm, nVidia is finally catching back up with ATi...not the other way around. :LOL:

It may just be that we are entering a second Golden Age of 3D Graphics
I'm curious, when do you consider the first Golden Age of 3D graphics was? :|
 
Oh, and I really did like all the bits about the manufacturing and dust and such. Very interesting reading about stuff I didn't know much about. :)

Sorry, me last post makes it sound like I didn't like your piece and I think on the balance I actually DID like it so felt it best to post up about it. ;)
 
JoshMST said:
Honestly Dave, I am not trying to misguide or push some kind of agenda, I am just working off the best information I could dig up and glean from anyone I could.

I know - I'm just feeling you out for info and also trying to steer your thinking. You've already mentioned some clues in this thread that I'm not sure whether you've put all together yet.
 
JoshMST,

Here a few major problems with the article :

The State of 3D October said:
"These forward looking thinkers decided on the “wide and shallowâ€￾ philosophy of pipeline design. The R300 had 8 parallel pixel pipelines that were floating point throughout (no internal integer support).

Take a look at this thread (over a year old) :

http://www.beyond3d.com/forum/viewtopic.php?t=8005&highlight=pipelines

http://www.beyond3d.com/reviews/ati/r420_x800/index.php?p=8

As you can see not really 'Shallow and wide'.

The State of 3D October said:

ATI has continually improved the R3x0 architecture throughout its lifetime. The latest version is the R420/R423, which expands the architecture to include PS 2.0b functionality (much more programmable and flexible than the original PS 2.0 specification). They have also increased the performance and functionality of the vertex shader units, so that a form of geometry instancing can be done even without full VS 3.0 functionality.

Look at the I've put in bold - you've stated that 'geometry instancing' was created for the X800, where it is now well known that it was designed in the R300 (9700) and all cards since (It's just resently been enabled in currect drivers for all cards including the R3X0 based ones).


This article reads as if you have skimed a few forums/online articles/PR documents and sort of 'sumed it up' with a bias toward Nvidia.

Also way to much speculation and opinion and no research or evidance to back up you claims!

I think you need to read and research a lot more about 3D archtechtures and the related 'design decisions' behind them before writing another 'The State of 3D' article!
 
PeterAce. While the items you posted about may be incorrect I fail to see how that would show an Nvidia bias. I didn't read the article so maybe that's why I'm missing something.
 
Status
Not open for further replies.
Back
Top