GeForce 5, or something else?

[Jason Cross]

NVIDIA is in the middle of deciding how they're going to name NV30. The discussion is whether they should call it GeForce 5 and take advantage of the brand equity in "GeForce," or name it something else entirely to promote the idea that this product is something amazingly new and different and not just "the same thing but moreso."

What do you guys think? GeForce 5, or something else? What about other NVIDIA naming practices (like naming different speed grades with numbers ala "GF4 Ti 4600, 4400, 4200, GF3 Ti 500, Ti 200, and so on)? How should they proceed?

While I can't describe NV30 to you, I should point out for the sake of this discussion that it has less in common with the GeForce 4 Ti cards than the GeForce 3 had with GF2 cards. It's a pretty drastic change/improvement.

[DavidCPA]

I'm in favor of a name change rather than the next numerical version.

-DavidCPA

[wumpus]

Sure, you can tell us about it. Those NDAs are just "suggestions".

I think a new name is in order, since the architecture is the first big change on the order of TNT2 -> GeForce.

GeForce 1 - T&L
GeForce 2 - faster, refined GeForce 1
GeForce 3 - faster, pixel shaders
GeForce 4 - faster, dual T&L, refined GeForce 3

I assume the new card is going to be a bigger change than any of these where.

[TimElhajj]

Yeah, it's pretty hard to get excited about another number. I think the branding focuse should be on the company name, not the product name.

[Anonymous]

Watch. They'll call it Geforce 5. You don't f*ck with a brand. :)

Peter

[Matthew Gallant]

nVidia GeForce DoomBlaster.

[xahlt]

Aren't they legally required to put in an Xtreme when they call it a Blaster of some sort?

[Anonymous]

They should call it "Zoltar" or "7-Zark-7."

[TimElhajj]

'They should call it "Zoltar" or "7-Zark-7."'

You gotta listen to name suggestions from a guy named, Gatchaman. :)

[Brad Grenz]

Didn't nVidia get the GeForce brand name from a contest they ran before the product debuted? Do I remember this, or am I halucinating it? I just know there's some guy out who got a free GeForce 1 card as payment for something a consulting firm would have charged millions for. I think I even entered myself. The name I came up with was "Prose". Oh, and I'd name the NV30 "Gravity".

[Anonymous]

They should call it "buy this P.O.S, which will not have programs effectively using its features until at least 2 years in the future (cause we need good quarterly revenue's) and the versions of it can be: FUCK the consumer 1, 2 etc.

btw when is this supposed to be released? I will need to upgrade my ge4 4600 immediatly...cant let theez n00bz b more l33t then me!!!!

[Chris]

Didn't Nvidia buy out 3dfx? They should use the Voodoo name for their MX cards to give a better idea of the differences in their two lines.

[DennyA]

Riva 129!

[balut]

TNT3: The Reckoning

- Balut

[wumpus]

Riva 129.. heh heh heh.. that's a good one.

[Wholly Schmidt]

GeForce5: Electric Boogaloo

Or simply label it (optimistically) "This Card Will Run Doom 3"

[Jason Cross]

Didn't nVidia get the GeForce brand name from a contest they ran before the product debuted? Do I remember this, or am I halucinating it? I just know there's some guy out who got a free GeForce 1 card as payment for something a consulting firm would have charged millions for. I think I even entered myself. The name I came up with was "Prose". Oh, and I'd name the NV30 "Gravity".

Yeah, there was a contest. And since the GeForce 256 was a "geometry accelerator" and was a "256-bit chip," GeForce 256 fit.

Now that they've moved beyond geometry acceleration as a selling point, and on to complex shaders, programmability, and so on, I personally think a name change is in order.

But when considering the whole market, OEMs and such, is the GeForce brand maybe what the majority of computer buyers are looking for? There's a lot of recognition for that name. Then again, a lot of general consumers recognize NVIDIA just as much. If they Dell slapped "NVIDIA xxxxx" on their product, would it be as strong a selling point?

Gravity is a pretty good name, by the way. And almost surely not something they'd get past legal. :?

[Alan Au]

Wasn't there some story about some Taiwanese chipmaker securing the rights to the name '586,' and then Intel promptly switching over to the 'Pentium' naming convention?

<insert Shakespeare quote about a 'rose' and 'any other name,' etc.>

- Alan

[Anonymous]

The nVidia nVigorator, of course!

[Wholly Schmidt]

"neVitably nFerior n the next V months"

[Brad Grenz]

Wasn't there some story about some Taiwanese chipmaker securing the rights to the name '586,' and then Intel promptly switching over to the 'Pentium' naming convention?

No, Intel tried to copywrite "586" and their application was denied, so they came up with Pentium.

[Anonymous]

The best joke on that is that Intel took the first 586 off the production line and did a simple calculation: 486+100. When the chip came back with 587, they decided to rename the damn thing Pentium.

Of course, you gotta remember the entire flawed Pentium recall at the time. It was really funny then.

[TimElhajj]

> you gotta remember the entire flawed Pentium recall at the time. It was really funny then.

Do I!

"Well, yes, the new chip is subject to errors. But they're able to calculate them very quickly!"

[Anonymous]

Just as an aside, Nvidia released the 30.82 drivers today, and I immediately got a 500-point jump in 3D2001SE from the 29.42 drivers. I'm running 1024x768x32 with 2xAA.

In other words: GET THESE DRIVERS NOW!

[Desslock]

>Intel tried to copywrite "586" and their application was denied, so they came up with Pentium

Trademark. I'm surprised Nvidia is even considering moving away from the Geforce name, given how attached these companies get to their brands.

[Anonymous]

It may be that they would consider a new name because they want consumers to associate with the Nvidia label rather than the card designation, especially with their diversification to mobo chipsets (which they slapped a "force" on the end of anyway). Then again, according to predictions the Radeon 9700 will have had the edge for several months, and consumers might just start associating GeForces with being second best. Personally, I think a lot of people get paid a lot of money to think about marketing this, but I'm sure it'll sell no matter how dumb the name might end up.

[Brad Grenz]

Trademark.

Ah, yup. I always mix those up.

[Anonymous]

Just as an aside, Nvidia released the 30.82 drivers today, and I immediately got a 500-point jump in 3D2001SE from the 29.42 drivers. I'm running 1024x768x32 with 2xAA.

In other words: GET THESE DRIVERS NOW!

They do speed up my GeForce3 immensely. I've no benchmarks to offer. But consider this: I have been playing Thief Gold for the last couple of weeks. I had to turn every graphic feature up to max just to knock the speed -down- enough to make it playable. :)

Peter

[Anonymous]

GeForce 10 from Navarone?? :)

[Anonymous]

". I had to turn every graphic feature up to max just to knock the speed -down- enough to make it playable."

That's kind of depressing. I guess we'll all have to keep multiple old PCs around the house to play old games:

* one from the early diskette era - DOS
* one from the early CD-ROM/Voodoo 1 era - DOS
* an early Windows 95 era machine with Voodoo 3
* a late Windows 95/Windows 98 TNT2 machine
etc. etc.

progress sucks

[Mike Cathcart]

"progress sucks"

Yeah. After reading that thread about System Shock the other day I decided to give it a try, but I only have a Win2K and a Debian box. Oh well, I was thinking about building a few extra systems for LAN games, anyway, so I guess this is an excuse to start on that.

Cool name.

[DennyA]

I'm still wondering why nobody's done a "PC emulator for the PC."

IE: Emulate an 8MB, VESA SVGA (probably a #9 card, remember them?), Soundblaster 16 with QEMM-style memory management.

[Mike Cathcart]

I'm still wondering why nobody's done a "PC emulator for the PC."

IE: Emulate an 8MB, VESA SVGA (probably a #9 card, remember them?), Soundblaster 16 with QEMM-style memory management.

This project (http://bochs.sourceforge.net/) actually emulates an entire IA32 environment so that you can run multiple operating systems simultaneously. Their screenshot page even has a shot of it playing Doom in DOS on a Linux box. They also have a few shots of it running Win9x with IE open, so I guess they're pretty serious. Not sure what the performance is like, though. I think I had read about this last year but never had a use for it. Maybe I'll give it a shot this weekend.

[Jason Cross]

Just as an aside, Nvidia released the 30.82 drivers today, and I immediately got a 500-point jump in 3D2001SE from the 29.42 drivers. I'm running 1024x768x32 with 2xAA.

In other words: GET THESE DRIVERS NOW!

How much is a 500-point jump?

I really hate how 3DMark gives you a score in obtuse "3DMarks". People have scores of 10,000 and they get all worked up over 100-point changes, not realizing that 100 points is 1% and within normal variations.

It also makes it unclear exactly what 3DMark is testing. Most of the hardcore game/hardware sites don't even seem to realize that the 3DMark score has nothing at all to do with those "fill rate" and "point sprite" and other synthetic tests. The 3DMarks are determined by adding together the frames per second scores of the game scenes alone. All that other stuff is just there to provide you with additional info if you want it.

It's a weighted sum, though. Which I'm not totally thrilled with.

Also, since the numbers are so big, it's impossible to put a 3DMarks score on the same graph as other benchmarks (which invariably measure straight FPS).

I really wish that MadOnion would change it for 3DM2002. Just make the 3DMarks score the average, not weighted, of the game scenes. People can use the breakdown if they want, or not.

It would keep people from going crazy over 1% changes because instead of the score jumping "100 points!! OMFG!!", it would jump 1 FPS or less. It would make it more clear that the score is derived only from the FPS of the "game scene" demos and not the extra synthetic tests. And it would allow all the hardware reviewers and stuff to put 3DMark scores together on graphs with other benchmark scores.

[Derek Smart [3000AD]]

While I can't describe NV30 to you, I should point out for the sake of this discussion that it has less in common with the GeForce 4 Ti cards than the GeForce 3 had with GF2 cards. It's a pretty drastic change/improvement.

I agree. And I already told them that calling it a geForce5 would be a big mistake.

Matrox did the right thing when they went with Parhelia because that architecture is nothing like the previous Millennium Gxxx series.

ATI can coast along on the Radeon 97xx brand name because the architecture of those cards is not such a drastic change from, say, the Radeon 8500. At least not from what I've seen.


I'm still wondering why nobody's done a "PC emulator for the PC."

IE: Emulate an 8MB, VESA SVGA (probably a #9 card, remember them?), Soundblaster 16 with QEMM-style memory management.

I think VMware (www.vmware.com) does something like that. Though I've never used it.

[Anonymous]

It's a 500-point jump on my old hardware, which is still running a 100-MHz buss and PC100 SDRAM.

Performance scales up big time with more advanced hardware, and a 500-point jump with a 440BX-based system would easily translate into a much larger jump with someone running a 1.6GHz system with DDR ram or RAMBUS.

[Jason Cross]

]
ATI can coast along on the Radeon 97xx brand name because the architecture of those cards is not such a drastic change from, say, the Radeon 8500. At least not from what I've seen.

Have you got a card now, Derek? I'm curious if they've fixed some of the specific problems you've complained about here (W buffer, fog table, etc).

The "architecture" may not be that drastically different from a silicon design standpoint, but the capabilities are drastically different. Still, the Radeon 9700 name is fine.

What is not fine is the stupid freakin' Radeon 9000. I mean, the card might be fine, but they totally borked the name. Okay, the Radeon 7xxx series were DirectX 7 parts. The Radeon 8xxx and FireGL 8xxxx cards were DirectX 8 cards. The Radeon 9700 is a DX9 card. But the Radeon 9000? Oh, it's a DX8 part.

And if 9000 is more than 8500, you'd think it would perform better. But it doesn't. It performs worse, for a variety of reasons.

That's fine because the price is also cheaper, and the card might even be a pretty good deal. But calling it the Radeon 9000 was a total mistake.

Way to pull an NVIDIA there, ATI. Nice to see that all the people publicly and loudly complaining about the misleading GeForce4 MX (including ATI!) didn't teach the boys in marketing a damn thing.

[Lee Johnson]

The problem with programs like VMWare is that their bag is hardware virtualization, not processor emulation. These take advantage of the VM86 mode that's been in Intel CPUs since the Pentium (actually since the 80386, though the VM86 mode on the 386 only supports a virtualized 8086). The software adds an emulation layer to handle those privileged instructions that aren't available in the VM86 mode, in addition to the code that provides the devices on the virtual PC.

(If I am incorrect in some of these details, I'm sure somebody will wander by to, um, gently correct me presently. :?)

Anyway, my point is that, for the most part, the code running on the virtual PC is running at the speed of the host--it isn't emulated, and the speed of execution cannot be controlled. That makes something like VMWare or plex86 an unsatisfactory solution for legacy gamers who want a machine that runs, say, no faster than a 486/50.

Bochs appears to emulate everything, including the CPU. It doesn't have to run on Intel hardware at all. That at least holds out the possibility that it could be tweaked to approximate different clock rates, depending on how hard it would be to fiddle around with the emulator's timing logic. It could well be that a modified version of Bochs is all we need to keep those old DOS games around and working indefinitely.

[Anonymous]

Bochs definitely does the trick for slowing down old DOS games. It makes my P3 650 run games like a 386-12. But it is a very faithful emulation of a PC with a VGS and a Soundblaster 16 sound card, so it is just the thing for running those old DOS games. And if it is running a game too fast (difficult to believe ATM) the speed is tweakable.

[Derek Smart [3000AD]]

Have you got a card now, Derek? I'm curious if they've fixed some of the specific problems you've complained about here (W buffer, fog table, etc).


Yeah and the fog table and other bugs were fixed in the most recent (6118) drivers.

The other two that I found (I really DO find it hard to believe that other devs don't find this stuff, because ATI confirms that I'm the first person to report them!!!*) have to do with the driver switching the buffer mode when doing text and lighting polys which are explicitly not supposed to be lit. This latter problem totally borked the specular lighting on my models when I turn that on in my rendering pipeline. They claim to have fixed them in the upcoming drivers; but I'm waiting for them to be released via the secure dev site.


The "architecture" may not be that drastically different from a silicon design standpoint, but the capabilities are drastically different. Still, the Radeon 9700 name is fine.

Agreed


What is not fine is the stupid freakin' Radeon 9000. I mean, the card might be fine, but they totally borked the name. Okay, the Radeon 7xxx series were DirectX 7 parts. The Radeon 8xxx and FireGL 8xxxx cards were DirectX 8 cards. The Radeon 9700 is a DX9 card. But the Radeon 9000? Oh, it's a DX8 part.

Yep. Well, nVidia pulled a similar stunt with the GF4MX didn't they?


And if 9000 is more than 8500, you'd think it would perform better. But it doesn't. It performs worse, for a variety of reasons.

That's fine because the price is also cheaper, and the card might even be a pretty good deal. But calling it the Radeon 9000 was a total mistake.

See GF4MX vs GF3 Ti :D


Way to pull an NVIDIA there, ATI. Nice to see that all the people publicly and loudly complaining about the misleading GeForce4 MX (including ATI!) didn't teach the boys in marketing a damn thing.

Of course not!! Now that would be sacriledge.

My biggest disappointment, to tell you the truth, is in the Matrox Pahrelia. I mean, WHAT were they thinking?

And to add insult to injury, well, here come the ladies over at Trident (http://www.gamersdepot.com/interviews/trident/xp4/001.htm)

At the end of the day, gamers win. Thats all I can say


*makes me feel like I was back a high school with some cheerleader telling me I was her first!

[Jason McCullough]

]
*makes me feel like I was back a high school with some cheerleader telling me I was her first!

I really need to macro this, but:

My eyes.....the goggles do nothing!

[Jason Cross]

]

Yep. Well, nVidia pulled a similar stunt with the GF4MX didn't they?


They did, and that was just as much bullshit. We got lots and lots of letters about that, and of course the displeasure was posted here.

It's funny how even ATI kind of poked NVIDIA for doing that, and then they turn around and do the same damn thing.

Rest assured, the bogus naming *WILL* be a prominent feature of my Radeon 9000 Pro review.

(and what's with the "Pro" anyway? I realise there's a faster model and a slower one, but the faster one is far from being aimed at the professional market or high-end anyone. It's like a $129 card!)

[Alan Au]

Maybe "Pro" is short for "profit?"

- Alan

[Doug Erickson]

What, exactly, is the Racdeon 9000? A clocked-up 8500?

[Xaroc]

Jason, I read on Anandtech that there will be a 9700 Pro a 9700 and a 9500 which would be similar to the GF4 4600, 4400, 4200 levels of performance. All of those will be feature complete just varying clock speeds.

-- Xaroc

[Jason Cross]

What, exactly, is the Racdeon 9000? A clocked-up 8500?

You'd think that based on the name, but NO.

Essentially, it's a Radeon 8500 with only one texture unit on each pipeline instead of two.

It's cheaper than an 8500, but it's also slower.


Jason, I read on Anandtech that there will be a 9700 Pro a 9700 and a 9500 which would be similar to the GF4 4600, 4400, 4200 levels of performance. All of those will be feature complete just varying clock speeds.

Well, yes and no. They've been thin on details, but a 9500 will probably have four pixel pipelines instead of eight and a 128-bit memory bus instead of 256-bit. Other than that it should be basically the same as a 9700. I would guess that, due to the improvements in their caching, bandwidth effeciency enhancement, and AA/anisotropic filtering improvements, it'll perform about like a Ti 4600. Maybe a little faster with 4X AA enabled. But it'll have all the DirectX 9 features, which is a huge bonus. Again - no real details out there yet, so it's hard to say for sure. I think it's supposed to hit just before the end of the year.

I'm still under embargo to review the Radeon 9700 so I can't give out specific benchmarks now, but let me tell you guys this: turn on AA and/or anisotropic filtering and this thing completely embarasses a Ti 4600. I haven't seen a performance delta like this is years.

[Met_K]

"*makes me feel like I was back a high school with some cheerleader telling me I was her first!"

Her first what? Enema? Oopsie Derek-poo, looks like you done got the wrong hole.

[Brad Grenz]

The "Pro" that ATI has started putting in their product names likely stems from the trouble they got into over clock speed differences between OEM and Retail version of the 8500.

The 9000s look real good next to the GF4 MX stuff they'll be competeing with, but the faster 8500 are so cheap these days the 9000 isn't worth it.

I'm actually pretty curious to see how the new Trident chip stacks up in the value market. I don't know where this DX9 stuff is coming from, though. The first time I ever saw DX9 and the XP4 in the same sentence was in the link of Shacknews.com to that Gamersdepot interview. I had always understood it to be a DX8 part. No way in hell that thing has the floating point pixel and 96/128 bit color or vertex 2.0.

[xahlt]

A nice factual article on the state of NV30 production at Anandtech:

http://www.anandtech.com/video/showdoc.html?i=1678

I'll be interested to see whether Nvidia agressively cuts prices on the new parts once production ramps up to surplus levels in late Winter or so. Theoretically the lead they are capitulating now could be regained when ATI moves to a lower micron process. It's like coming in for a pit stop, everybody's got to do it eventually.

[Jason Cross]

Brad - As you said, it's not a DX9 part (the Trident chip). I asked them about that in a phone interview a few weeks ago, and they said "the chip does everything required for DirectX 9."

Nice wording, I thought. So I said, "So it provides hardware acceleration for both vertex and pixel shaders 2.0 and displacement mapping?"

And after a little hemming and hawing, they confirmed it did NOT. It's a DX 8.1 part, and like every other DX 8.1 part they'll have DirectX 9 drivers. Whoop-de-doo. They said they'd have a fast $100 DX9 card for Xmas of '03, though.

Still, if the performance is anywhere near what they're promising (I betcha it's not, especially with AA enabled), it'll be a great budget card.

[Brad Grenz]

A nice factual article on the state of NV30 production at Anandtech

I can't believe they took so many pages to say the chip tapped out last weak. There was zero new information about the hardware. No final confirmation on memory configuration (my money's still on 128 bit DDR-II style memory) or performance speculation, other than to assure us it'll be the fasted thing when released. Anandtech keeps saying it's faster on paper, but there isn't even enough paper puplic to make such a judgement. nVidia talks a lot about the number of shader instructions per program, etc, but they don't bother to tell you it takes as many cycles to execute that many instructions on the NV30 as it would on a R300. It just couldn't all be containt in the same "program" on the ATI hardware. Bottom line is a shader program that complex would take so long to do on this generation's hardware as to render it completely worthless. By the time it matters the competition will have cuaght up or surpassed nVidia in that area.

Anandtech is really starting to annoy me. They've become such cheerleaders for Intel and nVidia it's aggrivating. This article is just as bad as the recent one about Intel's .09 micron process hich forgets to mention products based on it won't be available for over a year...

[Xaroc]

Anandtech is really starting to annoy me. They've become such cheerleaders for Intel and nVidia it's aggrivating. This article is just as bad as the recent one about Intel's .09 micron process hich forgets to mention products based on it won't be available for over a year...

I don't see the fan of Intel part. They picked a number of benchmarks in either the XP 2200 or P4 2.53 article that made the 2200 look to be within a few fps of the 2.53 whereas Tom's has mostly benchmarks showing the 2200 getting smoked by the 2.53. So I would say Anand is a bit more AMD biased than Intel biased. Also according to this:


By the end of this year Intel will have Prescott samples produced on the 90nm process sampling to partners, and by the time the second half of 2003 comes around Intel will be shipping 90nm parts in volume.

He did mention that it would be about a year until the .09mm process would be available. And that is on the first page of the article.

The issue is Intel is really back in the lead now no matter how you slice it. My next upgrade is going to be an intel. I like AMD and have built 4 of their systems for myself and family members but around August 25th I am going to be able to pickup a p4 2.26 for around $150 and overclock it to probably 2.8+ GHz with the retail heatsink. Nothing you can get from AMD can keep up with that plus I don't have to deal with massive heat issues or VIA if I go with a P4. I am not going to stay loyal to a particular company if another offers a better solution, like ATI over Nvida. I will see who comes out on top and get the reasonably priced card from one or the other.

-- Xaroc

[Derek Smart [3000AD]]

What, exactly, is the Racdeon 9000? A clocked-up 8500?

You'd think that based on the name, but NO.

Essentially, it's a Radeon 8500 with only one texture unit on each pipeline instead of two.

It's cheaper than an 8500, but it's also slower.

And they only did that because they want the 9xxx series of cards to be Direct9 compliant parts - hence this continued foolishness. Why can't everything be as simple as Matrox's naming conventions? :D

WARNING!!! Pre-requisite pimpage follows. You HAVE been warned.

Here is a shot from BCG (http://www.3000ad.com/pics/bcg/photo256.jpg) running on 9700 PRO with my pixel shader engine. That station closeup is running a glow map+specular map+detail map in a single pass. Pretty nifty aye?

This one (http://www.3000ad.com/pics/bcg/photo237.jpg) and this one (http://www.3000ad.com/pics/bcg/photo239.jpg) are shots of an infantry marine (totting a hernia-inducing GLE22 grenade launcher) and team of recon and elite force marines riding around in an assault vehicle. All tru-formed (http://www.3000ad.com/pics/bcg/photo251.jpg).

Enjoy!

[Jason Becker]

"turn on AA and/or anisotropic filtering and this thing completely embarasses a Ti 4600. I haven't seen a performance delta like this is years."

So what is it about the 9700 design that does this? AA has been a buzzword for several years but its always involved compramises if you wanted to use it. The 9700 sounds like the first card that can actually do it without tanking the framrate.

[Brad Grenz]

For one the 9700 has an assload more memory bandwidth than anything else which is usually the limiting factor on other tech. Also they're employing an intelligent multisampling method that is far more efficient than a standard supersampling technique.


He did mention that it would be about a year until the .09mm process would be available. And that is on the first page of the article.

I must have blown past that part. I just remember the summary on the last page making it sound like product was just about upon us. By the time .09 is out AMD will be in full press with Hammer. Intel has so much money in fabrication it's hard for anyone to go toe to toe with them there, but that's always been the case (well, for the last decade at least), but that didn't stop AMD from blowing past them with Athlon.