Slashdot Mirror

Anyone care to comment where a card like this Geforce will be REQUIRED?

Except for scientific aplications and video work
Nope, that's what the quadro fx line is for
http://www.nvidia.com/page/quadrofx_family.html

Yes, a grand is a lot of money for a video card. Especially for an end user. But there are already games that will take advantage of this. (For example, Doom III's ULTRA setting requires a video card with 512M of ram).

This card is also useful for those who want a card with the memory of a high-end Quadro (workstation graphics card), but don't want to spend $1700-2200.

Yeh, but running a few in parallel should help some. I think grandparent post is on the money with scientific visualisation, but I also think they'll be used for GPGPU-style processing on larger datasets. Nvidia seems to have similar ideas...

But what's especially cool is the chapter is available for free download on the GPU Gems 2 site.

There's a shortcut to a sample for this over at NVIDIA's develope r website. I looked into this briefly, and I was pretty impressed with the performance... The GPU always outperforms the CPU, and in most cases the improvement is a factor of 1.5-2.0.

Not as high as I would expect, but I expect this will increase over time.

While DSPs would certainly be faster for graphics than a regular CPU, you just can't beat a special-purpose pipeline, designed specifically for graphics.

True, but only if you have significant resources to invest in design and development and can amortise that huge up-front cost over many units.

I'd like to leverage existing, off-the-shelf general purpose design as much as possible to reduce the initial investment, particularly since the card is not planned to be a top performer anyway.

I've worked as a device driver writer on dozens of different image processing and graphics architectures (inhouse, not commercial) and it's really struck me that complex, special purpose, "high performance" hardware is vastly overrated. My experience has been that hardware engineers tend to over estimate the performance of special purpose hardware when compared to mass market processors, mainly because special purpose hardware is usually at least a generation behind the mass market general purpose devices.

Without exception the fast special purpose hardware turned out to be slow in practice and hard to work with due to many factors including the inability to handle corner cases (e.g. overflow or divide by zero), the inability to express data in the appropriate form (e.g. FP v. 16 bit integer), inability to work efficiently with software for functions not implementable in hardware (e.g. can't access critical registers efficiently), inability to scale (e.g. limited addressing, limited word size and single threaded) and bottlenecks in key places, particularly in getting data in and out. The mass market graphics cards get around these problems by throwing huge resources at it (and even then still tend to be buggy) but an OSS project can't afford to do that.

It's hard to put into words but the simple, fast, general purpose architectures with clean designs were a breath of fresh air and definitely gave the best overall result, including performance, when combined with decent software.

A bit like RISC and CISC in CPU's. Simple and fast, not complex and hardwired.

Display cards have historically tended from dedicated hardware pipelines to GPU's. The complexity and variety of problems are graphics subsystem must handle is extraordinary. I can tell you now that as a device driver developer flexibility and ease of programming is what's important, not getting speed in some obscure function I'm probably not going to use because it's missing something critical like overflow handling.

I don't want to over-emphasise these points but the hardware developers definitely should do a cost benefit analysis on alternative architectural approaches and talk to the software people about it.

---

Are you a creator or a consumer?

There's actually a staggering amount of performance to be gained out of doing stuff on the GPU... Anything that can be thought of as a massively parallel application will work pretty well there.

For example, FFTs can be done on the GPU at roughly 1.5-2x the speed of doing them on the CPU. (Of course, doing them on an FPGA will still trounce the GPU performance by--I believe--a factor of ~8.) But given a few more GPU generations, I believe this margin will be decreased signficantly.

There's a demonstration and whitepaper on this over here.

The Quadro FX 4400 has 512MB of RAM. Sun has one that has a gig of RAM but is really really specialized to the V800 server. However, 3dlabs has probably been there the longest, although I'm sure there are some more specialized ones out there.

Who cares? Modern graphics cards are capable of (sorry it's a PDF, it was all I could find) 40 GFLOPS. That's not even in SLI mode, which actually does push you to about a 98% over a single card (in terms of raw processing power).

Why would you buy a 96-CPU setup when you could buy a 6-GPU setup and match the same theoretical performance? (All jokes aside about the costs being roughly equivalent, they're nowhere near the same.) 6 top of the line 6800s would run you about $3600. Even if you added top of the line parts for the rest of the system, you'd be looking at about $1600 per system. Add $0 for the linux distribution to drive the whole thing, and you're at a grand total of $10K.

I'm not impressed.

The last time I updated my debian boxes the the latest, it wasn't quite that easy. The last time I updated my Windows box it went like this:

1. wget http://download.nvidia.com/Windows/71.89/71.89_win 2kxp_english.exe
2. 71.89_win2kxp_english.exe
3. Restart Windows

I haven't done #3 yet, nor have I waxed my xfce session and removed/re-inserted the new nvidia.o on the sarge laptop. I have too much crap open.

Oh wow! What a great idea! Why hasn't anyone thought of this before?!

Tell me, this great idea of yours... Would it look anything like this?

"The NVIDIA kernel module has a kernel interface layer which must be compiled specifically for the configuration and version of the kernel you are running. "

For the win.

64-bit AMD Opteron workstations were available two months before the G5 Powermacs were announced. If you call the G5 Powermacs "workstations" (dual processor, PCI-X), then AMD was first. If you call G5's "desktops" (no EEC memory, no workstation-class graphics cards), then PPC was first (before the Athlon 64 three months later).

2. I want a quiet system, and that means no fans. A high-end graphics board that sucks energy and produces heat is not just a neutral "I don't need it", it's a negative "I don't WANT it".

C'mon. A new high-end computer should have PCI Express. The Radeon 9600 is last generation's mid-range graphics. It is outperformed by this generation's low-end (GeForce 6200 and Radeon X700). Only the ultra high end cards need loud fans.

Don't forget, with that nVidia card, you can turn on hardware accelerated Window transparency (for both Windows 98 and XP). Its under "effects" in the nView configuration panel.
See the nView desktop manager guide
I was blown away when I found out about it.
Shame they bury it so deep.

Try the whitepapers here.

They have very good info, and little-to-no marketting BS.

From the article:

From what I have heard, more than a few games realize no FPS gains at all from the addition of a second video card, which essentially means they are not compatible. The games should run properly, but without two video cards, what's the point?

From NVIDIA's SLI FAQ :

Will my game or application just run on NVIDIA's SLI technology?

Yes.
Developers are not required to make changes to make their application work on NVIDIA's SLI solution. In fact, developers are not even required to make changes to enable the speed-boost available on a multi-GPU system.

What do I, as a developer, have to do to accelerate my application on a multi-GPU system?

Nothing:
NVIDIA's SLI technology accelerates applications automatically. The same guidelines as for maximizing performance on single-GPU systems apply. The NVIDIA GPU Programming Guide (available on developer.nvidia.com) discusses these guidelines; it also includes a chapter to specifically discuss multi-GPU considerations.

Link to the official press release, no need to guess further.

The motherboard mfgrs weren't getting enough positive feedback from it, it was expensive (the Dolby license was not the only reason), and Creative bought Sensura, whose technology it used.

Sure, you can come up with all the excuses you want for why soundstorm died, but we all know the real reason is this.

Yes. The nForce4 chipsets offload most TCP/IP processing and firewall from the main CPU.

If you go with a Athlon64 Socket 939 nForce4 board, you get PCI Express, lower power consumption, a ton of great features, good Linux support, and plug-compatible dual core upgrades down the road. Intel's offerings just seem anemic by comparison.

(Personally, I'd also do an NVIDIA graphics board for the excellent Linux driver support. And no, I don't work for NVIDIA, I'm just a satisfied customer.)

It's interesting to that a number of manufacturers are putting out glowing press releases about their involvement with this, including Transmeta. There's a bit more information on Transmeta's site about the actual specs of the wearable system itself.

And, yes, it can run Linux.

Ohhh... so it's kind of like Linux?

No, cause Linux has driver support

That's why I love nVidia. Their Linux driver works fantastic with my 6800GT and Doom 3!

"nView also provides increased efficiency on a single monitor by enabling multiple Windows desktops, quicker access to hidden windows with transparency and window rollups, and hotkeys for access to all nView functions. nView provides a quick and easy way for you to manage multiple Windows desktops, thereby increasing your efficiency and enabling you to see what you've been missing."

nView Multi Display Technology

I came across this while playing with my video card settings the other day. Some pretty neat options buried in there.

Perhaps. But I think another possibility is that the nForce3 chipset was not meant for "budget/mainstream" users, but for "enthusiasts." As we all know, enthusiasts don't want integrated graphics that share memory with the system.

The nForce4 chipset, on the other hand, does look like it's aimed at budget/mainstream users as well as enthusiasts. But with PCI Express and TurboCache, NVIDIA might have a cheap solution that's better than integrated graphics.

PCI Express x16 has more bandwidth than AGP (4 GB/s upstream and downstream) and allows writes directly from the GPU to system RAM. This allows a non-integrated graphics card to share memory with the system without the huge performance hit that AGP would have caused.

Instead of integrated graphics, maybe NVIDIA is planning to "bundle" their cheap TurboCache cards with nForce4 motherboards. That seems cool to me.

nVidia has a very nice treat: Unified driver architecture, which means drivers are the same for all GeForces and TNTs, from the original TNT up to the latest FX offerings. In practice, it means more or less they bundle all drivers in a single download (a 8mb binary driver? sheeze) - but they work just fine. This is for all platforms too; you can get them here.

As for OSS drivers, both X.Org and XFree86 include a basic TNT/GeForce driver with some degree of 2D acceleration but no OpenGL support.

ATI nVidia

That wasn't hard.

1. Video Cards, and
2. Games

Heard of the nForce4?

Of course, the submitter's reference to the Celeron was a joke (Turion = Asparagus, Celeron = Celery). I mostly agree that AMD has kicked Intel's arse in budget CPUs in the past. But I don't think AMD currently beats Intel in every budget segment.

Consider Anandtech's conclusion from a Sempron vs. Celeron test they did last July

Also consider current pricing (from Newegg) for the Sempron and the Celeron D. In Anandtech's benchmark results, the Socket 754-based Sempron 3100+ did beat the Socket 478-based Celeron D 335 in most of the bechmarks that count. However, the Sempron 3100+ costs $123/$108 (retail/OEM) while the Celeron D 335 costs $109/$89. The benchmarks also showed that the Socket A-based Sempron 2800+ ($109/$99) was about even with the Celeron D 335, but would you choose an aging Socket A platform (PCI, AGP, IDE, 333MHz FSB) over a modern platform (PCI Express, SATA, 533/800MHz FSB) that you can get with the Celeron D?

Of course, we're talking about building our own desktops, which is very different from what the big-name computer makers offer. Us home builders would choose nForce or 915 chipsets for Sempron and Celeron D CPUs. HP and Dell are more likely to offer VIA/SiS/ALi chipsets for Semprons and 865/845 chipsets for Celeron D. Ugh.

If I was building a budget desktop, I would choose a Socket 775-based Celeron D over a Socket A or Socket 754-based Sempron. I value the whole platform just as much (if not more) than the CPU itself. If AMD made a Socket 939-based Sempron, I'd reconsider.

I'll be interested to see how this unfortunately named "Turion" chip compares to the PentiumM.

Back to the article's topic (notebook CPUs), it looks like AMD will not have an answer to Intel's Celeron M. The Celeron M is based on the Pentium M core and performs almost equivalently clock-for-clock in Tom's Hardware benchmarks. Also note that Intel's Sonoma platform (533MHz bus, PCI Express, DDR2, GMA900 graphics, HD Audio) is about to be lauched.

Gee whiz folks, we've got UT2004 with a NATIVE PORT, we've got Doom 3 with a NATIVE PORT (excellent port thank you ID), HL2 plays under Cedega (you need a brute of a CPU tho.) Lots of games with nice installers: Check out http://liflg.org/ for installers for the best games.

People cry that it's too hard to get your video to work; I'm really sorry you were too cheap to get the GeForce Go 5200 and got the Intel "Extreme" integrated graphics and now your pixel shader games look crappy. Nevertheless, UT2004 and the UT2004 DEMO play under linux with the DRI drivers.

Stop complaining, Loki is gone, but http://icculus.org/ is still around, and several of those guys WORKED for loki. If you want it EASY, you want GAMES then either USE WINDOWS or buy an X-Box. If you prefer Linux and are willing to expend the time and energy (and reap the rewards of what you learn) then USE LINUX and play the games that work well, there are a bloody AWFUL lot of games that work, work well and aren't that difficult to set up.

Take a little time, subscribe to http://www.transgaming.com/, make a little donation to http://liflg.org/, buy products from http://www.nvidia.com/ and shut up & enjoy the games!

Cedega is a non-free version of wine with directx capabilities. You can browse their supported games here.

Of course not all games now-a-days require wine or cedega in order to run on linux. Games like unreal tournament and doom III include fully functional linux versions.

There are several open source games developed for but not limited to linux. torcs, flightgear, tuxracer are some examples.

Projects like libsdl are making cross-platform game development easier.

Probably the biggest problem you'll encounter is building drivers for your video card. I've heard it argued both ways but as I understand it, both nvidia and ati drivers are ass-pains in linux. Nvidia's drivers are free as in beer, not speech. If you don't really care about free-software principles and philosophy then this is not a problem for you. ATI's drivers I understand to perform less than ideally. If you haven't already purchased your video card, I would encourage you to do extensive research beforehand.

In reality, linux distributions have few differences. Any recent, major distribution should be able to accomodate gameplay. I myself use debian unstable for amd64.

As far as performance, it really boils down to hardware. My advice is to install the linux distribution of your choice. Once you get glxgears to run, give ut2004demo a try, and if you like the way it works, then stick with linux.

I think we still are, but I don't think this product is targeted at us. NVIDIA's web page for GeForce 6200 with TurboCache technology describes it as a product "for entry-level PCs." Also, I think this product (when paired with nForce4) might be NVIDIA's low-cost PCI Express answer to Intel's integrated GMA 900 graphics.

Ever since Intel's 810 chipset, almost all "entry-level" platforms have included integrated graphics with "shared" memory architectures. Intel has dominated this category (in sales) and this continues with their current PCI Express platforms with integrated GMA 900 graphics. Note that GMA 900 is (barely) DirectX 9 compatible.

NVIDIA's nForce4 chipsets with PCI Express, which are just starting appear in stores, don't have a version with integrated graphics. If NVIDIA can offer to OEMs an nForce4 (for Athlon 64 and Sempron) + NVIDIA 6200 w/TurboCache bundle that's competitively priced with Intel's 915G chipset, then we might see more big-name computer makers selling low-cost Athlon 64/Sempron systems.

I think we still are, but I don't think this product is targeted at us. NVIDIA's web page for GeForce 6200 with TurboCache technology describes it as a product "for entry-level PCs." Also, I think this product (when paired with nForce4) might be NVIDIA's low-cost PCI Express answer to Intel's integrated GMA 900 graphics.

Ever since Intel's 810 chipset, almost all "entry-level" platforms have included integrated graphics with "shared" memory architectures. Intel has dominated this category (in sales) and this continues with their current PCI Express platforms with integrated GMA 900 graphics. Note that GMA 900 is (barely) DirectX 9 compatible.

NVIDIA's nForce4 chipsets with PCI Express, which are just starting appear in stores, don't have a version with integrated graphics. If NVIDIA can offer to OEMs an nForce4 (for Athlon 64 and Sempron) + NVIDIA 6200 w/TurboCache bundle that's competitively priced with Intel's 915G chipset, then we might see more big-name computer makers selling low-cost Athlon 64/Sempron systems.

I think we still are, but I don't think this product is targeted at us. NVIDIA's web page for GeForce 6200 with TurboCache technology describes it as a product "for entry-level PCs." Also, I think this product (when paired with nForce4) might be NVIDIA's low-cost PCI Express answer to Intel's integrated GMA 900 graphics.

Ever since Intel's 810 chipset, almost all "entry-level" platforms have included integrated graphics with "shared" memory architectures. Intel has dominated this category (in sales) and this continues with their current PCI Express platforms with integrated GMA 900 graphics. Note that GMA 900 is (barely) DirectX 9 compatible.

NVIDIA's nForce4 chipsets with PCI Express, which are just starting appear in stores, don't have a version with integrated graphics. If NVIDIA can offer to OEMs an nForce4 (for Athlon 64 and Sempron) + NVIDIA 6200 w/TurboCache bundle that's competitively priced with Intel's 915G chipset, then we might see more big-name computer makers selling low-cost Athlon 64/Sempron systems.

[NEOtaku17] This completely debunks the myth that is this Slashdot article.

FOR IMMEDIATE RELEASE:

TOKYO and SANTA CLARA, CA--DECEMBER 7, 2004--Sony Computer Entertainment Inc. (SCEI) and NVIDIA Corporation (Nasdaq: NVDA) today announced that the companies have been collaborating on bringing advanced graphics technology and computer entertainment technology to SCEI's highly anticipated next-generation computer entertainment system. Both companies are jointly developing a custom graphics processing unit (GPU) incorporating NVIDIA's next-generation GeForce(TM) and SCEI's system solutions for next-generation computer entertainment systems featuring the Cell* processor.

This collaboration is made under a broad, multi-year, royalty-bearing agreement. The powerful custom GPU will be the graphics and image processing foundation for a broad range of applications from computer entertainment to broadband applications. The agreement will encompass future Sony digital consumer electronics products.

"In the future, the experience of computer entertainment systems and broadband-ready PCs will be fused together to generate and transfer multi-streams of rich content simultaneously. In this sense, we have found the best way to integrate the state-of-the-art technologies from NVIDIA and SCEI," said Ken Kutaragi, executive deputy president and COO, Sony Corporation, and president and Group CEO, Sony Computer Entertainment Inc. "Our collaboration includes not only the chip development but also a variety of graphics development tools and middleware, essential for efficient content creation."

"We are thrilled to partner with Sony Computer Entertainment to build what will certainly be one of the most important computer entertainment and digital media platforms of the twenty-first century," added Jen-Hsun Huang, president and CEO, NVIDIA. "Over the past two years NVIDIA has worked closely with Sony Computer Entertainment on their next-generation computer entertainment system. In parallel, we have been designing our next-generation GeForce GPU. The combination of the revolutionary Cell processor and NVIDIA's graphics technologies will enable the creation of breathtaking imagery that will surprise and captivate consumers."

The custom GPU will be manufactured at Sony Group's Nagasaki Fab2 as well as OTSS (joint fabrication facility of Toshiba and Sony).

This seems a bit unlikely given the approach Sony took with the PS1 and PS2, but there's a press release to back this up on nVidia's site.

Other than the fact that this is old news. I would have figured that the focus would be more on the new nforce4 chipset http://www.nvidia.com/page/nforce4_family.html familiy. There are three board types in this family The Nforce4 Standard, the Nforce4 Ultra, and the Nforce SLI. As a matter of fact Asus is releaseing an sli board based on this right now called the A8N-SLI with a slew of added features that you could expect out of and asus board including dual gigabit ethernet ports! Why the via board is even being covered is beyond me the nforce is a much more better chipset. Here is a [H]ardOCP benchmark page here http://www.hardocp.com/article.html?art=Njk2. Enjoy. ;)

--
So really, who is hotter? Alley or Alley's sister?

High-end Quadros have 512 MB RAM. plus, they're dirt cheap ;)

It is worth noting that NVIDIA will be bringing SLI to the Intel platform according to this press release:

http://www.nvidia.com/object/IO_17070.html

I'm looking forward to a P4 NForce board.

1. SLI graphics (2 graphics cards powering 1 monitor)
2. 3Gb/s SATA (Intel uses 1.5Gb/s SATA)
3. NVIDIA ActiveArmor (firewall)

Unfortunately, like you said, it seems like NVIDIA dumped SoundStorm. Meanwhile, Intel added Intel HD Audio, which apparently encodes Dolby Digital 5.1 in real time.

1. SLI graphics (2 graphics cards powering 1 monitor)
2. 3Gb/s SATA (Intel uses 1.5Gb/s SATA)
3. NVIDIA ActiveArmor (firewall)

Unfortunately, like you said, it seems like NVIDIA dumped SoundStorm. Meanwhile, Intel added Intel HD Audio, which apparently encodes Dolby Digital 5.1 in real time.

1. SLI graphics (2 graphics cards powering 1 monitor)
2. 3Gb/s SATA (Intel uses 1.5Gb/s SATA)
3. NVIDIA ActiveArmor (firewall)

Unfortunately, like you said, it seems like NVIDIA dumped SoundStorm. Meanwhile, Intel added Intel HD Audio, which apparently encodes Dolby Digital 5.1 in real time.

1. SLI graphics (2 graphics cards powering 1 monitor)
2. 3Gb/s SATA (Intel uses 1.5Gb/s SATA)
3. NVIDIA ActiveArmor (firewall)

Unfortunately, like you said, it seems like NVIDIA dumped SoundStorm. Meanwhile, Intel added Intel HD Audio, which apparently encodes Dolby Digital 5.1 in real time.

BZZZZZZT!!! Wrong answer.

You are wrong. The NVidia Unified Driver Architecture (UDA) shares 95% of its code base between all supported operating systems. That other 5% obviously needs to be different because each OS has a different way to interface with the kernel/graphics system.

The price is free when you have an Nvidia GPU, which I'm sure a lot of you do.

Click here to get an Nvidia only free(beer) version. Their site seems to be down at the moment, which is odd for such a large company, but when it comes back up, you can get it from there. There are many other cool programs you can get for free if you have an Nvidia card while you are there.

[root@rsd800fc3 ~]# modprobe nvidia
[root@rsd800fc3 ~]# cp -a /dev/nvidia* /etc/udev/devices
[root@rsd800fc3 ~]# chown root.root /etc/udev/devices/nvidia*

And I'm also certian that the US didn't just complete the first non-government manned space flight and doesn't have billions of dollars going to develop private space flight.

Give me a break.

China is emerging as an ecenomic powerhouse, and it looks like it will continue down that path, provided their government doesn't screw up. However please don't pretend like all good things come from China. I gave just a small list of the US companies that produce advanced hardware, including what drives almost all the devices you listed. Your MP3 player may be built in China but it's usually using TI DSPs and AD opamps.

You know it's perfectly possible for China AND the US to be economic powers, and for both to benefit from trade with each other.

Simply not true. Unlike ATI, NVIDIA has provided excellent Linux driver support for some time. Note that in the release notes for the latest release, one item is "Fixed problem in GLSL with shadow2DProj.". This indicates support for a bleeding-edge OGL feature.

True, Doom III runs somewhat slower under Linux. However, it's hard to say why at this point. There are many more variables than just the graphics drivers. I'm sure NVIDIA (which seems quite committed to Linux) will work hard to squeeze every ounce of performance out of its hardware on that platform.