Slashdot Mirror

Anyone know how this compares to a GTS 250-based card? As those are also DX10-compliant and can be easily found for around $120, I'm not sure what the value of this new model is... beyond the psychological impact of hitting the magic $99 price point, of course.

(More pictures, including the obligatory model to hang off the computer like it was a sugar daddy.)

(More pictures, including the obligatory model to hang off the computer like it was a sugar daddy.)

You might have a little bit more look into the fermi archetecture. Based on http://www.nvidia.com/content/PDF/fermi_white_papers/NVIDIAFermiArchitectureWhitepaper.pdf nvidia's white paper, and assuming a clock speed of 600mhz, i.e. in line with a GTX 280, they are looking at 1.5 Tflops of DOUBLE precision computing power. Nvidia is making a hell of a push at multi-threaded supercomputing. Not to mention some crazy cache sizes.

Your sarcasm might be true for gaming, but Linux rules on scientific / number crunching applications, and that is what Tesla is built for. From Tesla Tech Specs, Windows XP is not even supported on the higher end S870 model.

Nvidia has had a page up for a while on this. Most of these units use small desktop motherboards, have limited ram and IO capability, and lots of GPU. These are poor designs for many calculations. These guys have a dual socket, full server class motherboard with up to 144 GB of ECC DDR3 RAM, as well as putting more than 500 MB/s into their local disk IO channel with up to 32 2.5 inch SATA or SAS drives, in a single quiet deskside chassis.

Asus wasn't the first. They are about a year late to the party.

While this sort of machine is useful (I just built one for quantum Monte Carlo calculations 6 months ago) it is hardly news. NVIDIA has been pushing this sort of machine since the launch of the Tesla. In fact, they have had a parts list on their website for some time telling exactly what is needed to put together a computer with 4 C1060's. This is not even the first commercial offering of this nature, with companies like appro and microway having similar products for at least a year (see nvidia) for a more complete list.

While this sort of machine is useful (I just built one for quantum Monte Carlo calculations 6 months ago) it is hardly news. NVIDIA has been pushing this sort of machine since the launch of the Tesla. In fact, they have had a parts list on their website for some time telling exactly what is needed to put together a computer with 4 C1060's. This is not even the first commercial offering of this nature, with companies like appro and microway having similar products for at least a year (see nvidia) for a more complete list.

and have up to four teraflops ... http://www.nvidia.com/object/tesla_supercomputer_wtb.html

The Tesla c1060 processor boards sound like a very efficient way of packing in compute power, but unless they're neglecting to mention it, the 4GB of GDDR3 RAM each has on board has no error correction. Given the rates of correctable errors observed e.g. here, I could never recommend using it for computing simulations that matter. A flipped bit in a floating point number can have a disproportionate affect on the outcome of calculations that rely upon it, and short of running the whole simulation a second or third time, one couldn't be confident that such an error did not occur.

Large compute-intensive simulations can take weeks, and are used to justify engineering and business decisions that involve the disposition of large amounts of money and other resources — it is important that the computational part of the process can be relied upon.

Which is why the upcoming NVIDIA "Fermi" GPU based boards will support 4GB of ECC memory. Also, they'll have about 2 TFLOPS of single-precision power, and you can stack 4 of them in a box = 8 TFLOPS beside your desk.

I can't wait until the US government starts banning these things because they could be used by terrorists to design nuclear weapons or something. 8)

The Tesla c1060 processor boards sound like a very efficient way of packing in compute power, but unless they're neglecting to mention it, the 4GB of GDDR3 RAM each has on board has no error correction. Given the rates of correctable errors observed e.g. here, I could never recommend using it for computing simulations that matter. A flipped bit in a floating point number can have a disproportionate affect on the outcome of calculations that rely upon it, and short of running the whole simulation a second or third time, one couldn't be confident that such an error did not occur.

Large compute-intensive simulations can take weeks, and are used to justify engineering and business decisions that involve the disposition of large amounts of money and other resources — it is important that the computational part of the process can be relied upon.

Well, in Gentoo's Portage, the nvidia-drivers ebuild states the sources as:
SRC_URI="x86? ( ftp://download.nvidia.com/XFree86/Linux-x86/${PV}/${X86_NV_PACKAGE}-pkg0.run )
                  amd64? ( ftp://download.nvidia.com/XFree86/Linux-x86_64/${PV}/${AMD64_NV_PACKAGE}-pkg2.run )
                  x86-fbsd? ( ftp://download.nvidia.com/freebsd/${PV}/${X86_FBSD_NV_PACKAGE}.tar.gz )"

Why waste the own mirrors' bandwidth on it? ^^

Well, in Gentoo's Portage, the nvidia-drivers ebuild states the sources as:
SRC_URI="x86? ( ftp://download.nvidia.com/XFree86/Linux-x86/${PV}/${X86_NV_PACKAGE}-pkg0.run )
                  amd64? ( ftp://download.nvidia.com/XFree86/Linux-x86_64/${PV}/${AMD64_NV_PACKAGE}-pkg2.run )
                  x86-fbsd? ( ftp://download.nvidia.com/freebsd/${PV}/${X86_FBSD_NV_PACKAGE}.tar.gz )"

Why waste the own mirrors' bandwidth on it? ^^

Well, in Gentoo's Portage, the nvidia-drivers ebuild states the sources as:
SRC_URI="x86? ( ftp://download.nvidia.com/XFree86/Linux-x86/${PV}/${X86_NV_PACKAGE}-pkg0.run )
                  amd64? ( ftp://download.nvidia.com/XFree86/Linux-x86_64/${PV}/${AMD64_NV_PACKAGE}-pkg2.run )
                  x86-fbsd? ( ftp://download.nvidia.com/freebsd/${PV}/${X86_FBSD_NV_PACKAGE}.tar.gz )"

Why waste the own mirrors' bandwidth on it? ^^

From the nvidia-installer man page:

              --update
                            Connect to the NVIDIA FTP server ’ ftp://download.nvidia.com ’ and determine the lat
                            est available driver version. If there is a more recent driver available,
                            automatically download and install it. Any other options given on the commandline
                            will be passed on to the downloaded driver package when installing it.

I wonder whats the use for 3D laptop, and if this works better than the existing tech?

NVIDIA 3D Vision is great with some games, but laptops aren't usually used for that and you would probably want atleast 17" screen if you'd get it for gaming. So whats the use?

In the last year I've worked on two different movies that were filmed stereoscopically. There were times where being able to play back a stereoscopic .mov would have been awesome.

I wonder whats the use for 3D laptop, and if this works better than the existing tech?

NVIDIA 3D Vision is great with some games, but laptops aren't usually used for that and you would probably want atleast 17" screen if you'd get it for gaming. So whats the use?

I am looking forward to how the Android mobile devices look over the next year, and especially for those including NVidia's Tegra APX chip line. They have been hyping it long enough as already integrated to work with the Android OS, I would hope some mobile device makers have jumped on board.

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

~Zwergin

Let me add to that list:

- A kick ass mascot that your girlfriend would like.

- Netgraph: try it and never look back.

One bad point that hasn't been mention: nvidia-freebsd-amd64's non-existance. This long-lasting feud seems to be resolved but Nvidia has yet to show the goodie.

...I happen to know of a school with a computer lab full of those Physx cards, and the majority of them have ATI/AMD cards for graphics. For them at least, this update renders all those standalone cards useless.

The article says this problem is in the new forceware drivers, not the standalone physx drivers. If the computer does not have an NVIDIA card, why would they be installing the forceware drivers?

I always found ATI cards work better for me (HTPC setups - I am not a gamer), however even I am not a "fanboi", so I can easily see Charlie's strong bias. It is not that he makes up facts, as far as I have seen, he bases his articles on information that turns out to be true or mostly true. However, he blows things way out of proportion, and his sarcastic style of writing is most definitely not proper for journalistic use.

In this case, he does have a good point. Go to http://www.nvidia.com/object/fermi_architecture.html and click on the announcement video. At around the middle, he says "I have one sitting right here. This, ladies and gentlemen, this puppy here is Fermi". Well, they now admit it wasn't. Yes, they probably had an engineering sample with cables coming out running somewhere in the back, but showing it as a polished product in their presentation is dishonest, since it just tries to make people believe the release is just around the corner. If they held up just the gpu and declared "this is fermi", it would be honest but would do anything about the impression that ATI has their latest generation out while nVidia is far from a release.

However, "paper launches" like these are common, and they often are dishonest - this is the corporate world after all, they will try and beat the competition with any means. So, while this is fine for a front page news for an anti-nVidia website, it is not an event of huge proportions. There is no point badmouthing Charlie, he did what he does well (bashing nVidia), he did it with accurate information, it is just that the issue is not really worth that much discussion. Of course Slashdot usually picks up articles that are MUCH, MUCH worse than this, so, we shouldn't be surprised this made front page ;)

There were a hand-full of issues behind that decision. One of them was that some GPGPU platforms fail silently, which, in practice, means that you start crunching numbers with less than the expected mantissa and therefore you get considerably larger rounding errors,. This is something that may bring disastrous results. Another issue is that even in some cases the announced double-precision support of some products was a bit flawed, as it failed to comply with IEEE 754, the standard for floating-point arithmetic. Although it didn't complied due to only a hand-full of issues, to rely on GPGPUs to crunch numbers when they don't conform to that standard would mean that someone would be forced to spend a considerable time formally checking what effects that non-compliance would have on the project being developed. That means that that would take precious man-hours from projects which may already be poorly manned, not to mention that that task would be rendered to waste as the next GPGPU generation would either fully support with IEEE 754 or, in the worst case scenario, fail to support it in some other aspect, which would mean that the poor chap assigned to verify the effects of the product's non-compliance would be forced to do everything from scratch, once again.

So, to sum things up, GPGPU's support for double-precision math is, in fact, great news. It means that everyone may have it's own personal vector-processing super-computer on his desktop. Heck, even on laptops. That may not mean much for the proverbial joe-sixpack (at least not beyond the "oohh... shiny graphics" side of things) but being able to crunch a lot more numbers on the same time frame means the world to anyone writing/using number-crunching software, which is a lot of people.

Onboard GPUs don't have PhysX unless they belong to the list here: http://www.nvidia.com/object/physx_gpus.html
Now don't get me wrong, I too think this is a shitty move, but I'd scream more for SLi Support on Intel Motherboards over this. IMO we're annoyed by the lesser evil here.

Nvidia says they support any 100+Hz CRT.

http://www.nvidia.com/object/3D_Vision_Requirements.html

Yes the 780g and later graphics chips are great for integrated graphics. I built my wife a system with a 780g and an AMD low power X2 cpu. She loves it because it is so quit and doesn't heat up the room like her old system did.
But ATI/AMD doesn't offer an Atom class system for netbooks and it doesn't offer an Atom chipset to compete with ION.
I am not so sure about flash hardware acceleration everywhere yet. This from June http://www.nvidia.com/object/io_1243934217700.html only talks about nVidia and I couldn't find anything about ATI but I might have missed it. Flash so isn't dieing. I wish that it would die but it is alive and well and thriving. YouTube and Hulu are prime examples of how Flash is alive and well. Would I like to see Flash replaced with an open standard? Yes but that has not happened yet. Theora with out native IE and Safari support is DOA.
Yes I know about OpenCL the problem is that CUDA is ahead of it. I want more support for OpenCL and less for CUDA.
Tegra is interesting but TI has a TI has OMAP which is very Tegra like as is the the Qualcomn Snapdragon.
nVidia sure isn't DOA and lots of people buy their cards and will for a long time.
The ZuneHD is using the Tegra and that alone will bring in a nice chunk of cash.
I have no real love for nVidia except that I have had a lot better luck with the nVidia graphics cards I have owned than the ATI cards. The exception is the Asus 780g motherboard in my wife's pc.

From the NVIDIA PhysX FAQ:

Can I use an NVIDIA GPU as a PhysX processor and a non-NVIDIA GPU for regular display graphics?
No. There are multiple technical connections between PhysX processing and graphics that require tight collaboration between the two technologies. To deliver a good experience for users, NVIDIA PhysX technology has been fully verified and enabled using only NVIDIA GPUs for graphics.

Or you could just buy them pre-configured...

Wouldn't an NVIDIA Tesla based system give you a lot more horsepower for a lot less money, and a lot lower power consumption?

Ah, here it is... http://www.nvidia.com/object/hybrid_sli.html

Im pretty sure that also got discontinued with the 9xxx generation of nVidia GPU's

>>Finally, we can have asynchronous GPU pairing?

I think NVIDIA has some sort of asymmetrical SLI mode available on its mobos with built-in video cards. It allows the weak built in card to help a little bit with the big video card installed in the main PCI-E slot.

IIRC, it gives a 10% boost or so to performance.

Ah, here it is...
http://www.nvidia.com/object/hybrid_sli.html

As for Nvidia, do they have any chips that fit in the price and power brackets of netbooks?

Hmm, apparently they do, in the form of their own ARM SoC...

You realise that nVidia has licensed the ARM Cortex A9 core that TFA is about and is shipping chips based on it, right?

No, but NVidia has gone ahead and integrated ARM.

Does ARM plan on integrated video along the lines of Nvidia and ION? http://www.nvidia.com/object/sff_ion.html

This all 3D-to-the-eyes is an old trick, but over the past year I've started to think that now theres actually good technology available for it.

I purchased myself the NVIDIA 3D Vision and played with it on various games. My favourite game for the past year has been left4dead and the 3D effect on it is really outstanding - everything looks so much scarier and you actually feel like being there.

The old cheap tricks are quite obsolote now as tech has improvent. But the future of gaming and movies surely is in this 3D and "be there" experience. Even MS and Sony have admitted that just pushing megapixels and polygony amount isn't the best thing, as they're at their maximum already anyway. We always see these things in movies, but the technology isn't really far from there now.

Now the only thing is about making it convenient for end users.

I love it when Apple fanboys ask for examples!

"It's funny because you did not provide even one example of how the PC provides more options than your Mac... something rather critical to your premise, no?"

Take a look at this:

Download.com > Educational Software > Science Software for windows
http://download.cnet.com/windows/educational-software/?tag=ltcol;nav

Download.com > Educational Software > Science Software for mac
http://download.cnet.com/mac/science-software/3150-2054_4-0.html

Look at the number of titles on the left. Compare the number of PC titles to Mac titles.

Then have a look here:

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

How many options for Mac can you find?

They already have (see also here). Also, check out OpenCL. OpenCL is the future, we've begun testing/working with it at my office for calculation engine. It's still in its' infancy, but it can only get better!

They already have (see also here). Also, check out OpenCL. OpenCL is the future, we've begun testing/working with it at my office for calculation engine. It's still in its' infancy, but it can only get better!

Look at nVidia's website. They already provide HPC solutions,

http://www.nvidia.com/object/tesla_computing_solutions.html
http://www.nvidia.com/object/preconfigured_clusters.html

http://www.nvidia.com/object/product_tesla_s1070_us.html

# of Tesla GPUs 4
# of Streaming Processor Cores 960 (240 per processor)
Frequency of processor cores 1.296 to 1.44 GHz
Single Precision floating point performance (peak) 3.73 to 4.14 TFlops
Double Precision floating point performance (peak) 311 to 345 GFlops
Floating Point Precision IEEE 754 single & double
Total Dedicated Memory 16 GB
Memory Interface 512-bit
Memory Bandwidth 408 GB/sec
Max Power Consumption 800 W
System Interface PCIe x16 or x8
Software Development Tools C-based CUDA Toolkit

Maybe point your quantum mechanics simulations people to HPC specific nVidia solutions, not commodity hardware not designed for HPC (ie. no RAM, limited cores, etc.)

Look at nVidia's website. They already provide HPC solutions,

http://www.nvidia.com/object/tesla_computing_solutions.html
http://www.nvidia.com/object/preconfigured_clusters.html

http://www.nvidia.com/object/product_tesla_s1070_us.html

# of Tesla GPUs 4
# of Streaming Processor Cores 960 (240 per processor)
Frequency of processor cores 1.296 to 1.44 GHz
Single Precision floating point performance (peak) 3.73 to 4.14 TFlops
Double Precision floating point performance (peak) 311 to 345 GFlops
Floating Point Precision IEEE 754 single & double
Total Dedicated Memory 16 GB
Memory Interface 512-bit
Memory Bandwidth 408 GB/sec
Max Power Consumption 800 W
System Interface PCIe x16 or x8
Software Development Tools C-based CUDA Toolkit

Maybe point your quantum mechanics simulations people to HPC specific nVidia solutions, not commodity hardware not designed for HPC (ie. no RAM, limited cores, etc.)

Look at nVidia's website. They already provide HPC solutions,

http://www.nvidia.com/object/tesla_computing_solutions.html
http://www.nvidia.com/object/preconfigured_clusters.html

http://www.nvidia.com/object/product_tesla_s1070_us.html

# of Tesla GPUs 4
# of Streaming Processor Cores 960 (240 per processor)
Frequency of processor cores 1.296 to 1.44 GHz
Single Precision floating point performance (peak) 3.73 to 4.14 TFlops
Double Precision floating point performance (peak) 311 to 345 GFlops
Floating Point Precision IEEE 754 single & double
Total Dedicated Memory 16 GB
Memory Interface 512-bit
Memory Bandwidth 408 GB/sec
Max Power Consumption 800 W
System Interface PCIe x16 or x8
Software Development Tools C-based CUDA Toolkit

Maybe point your quantum mechanics simulations people to HPC specific nVidia solutions, not commodity hardware not designed for HPC (ie. no RAM, limited cores, etc.)

The 120Hz display bit is here:

http://www.nvidia.com/object/3D_Vision_Overview.html

This requires an NVIDIA graphics card and a 120Hz display. Furthermore, it's damn cool!

without any real thought?
he is just saying "we hope to keep current trend for next six years" in a way, to get free press
in other words:
- there are some specific calculations, where current GPU is 50x faster than CPU
- suppose, that GPU computing speed will continue to grow by 50% annually
- then after 6 years, we can have GPU 50*1.5^6 = 570x faster than current CPU
see drawing board at http://blogs.nvidia.com/ntersect/2009/08/hot-chips-2009-keynote-by-jen-hsun-huang.html

You mean like the Tesla?

No, that won't do. The NVIDIA architecture (which is shared between Tesla and graphic cards) is 32-bit, meaning that it can only flat-address 4GB of RAM tops. The more sophisticated Tesla solutions are essentially built from clusters of Tesla cards, each with its own 4GB of RAM tops. Separate memory spaces means expensive memory transfers to share data between the cards, which is not an issue if you can get good domain decomposition, but is a BIG issue if you cannot.

The revolution for HPC on GPUs would be a 64-bit GPU architecture.

Proper support for doubles and possibly even long doubles would be a plus, for applications that need it.

You mean like the Tesla?

After reading most in this thread it is strange that no one mention the new NVIDIA Tegra that is on the way.

I am developing (just started) apps for Android but are waiting to buy a phone until I can get one with Tegra.

Ok, Tegra is not a Android only cpu (actually more than just a cpu too) but it is JUST what the Android need. Not an obscure manufacture but one of the biggest and most know and respected.
NVIDIA is great at reference designs and that will make it fast and easy (read cheap) to make an Android phone for any hardware manufacture. Samsung is saying that next year we will see sub $100 Android phones, and that can iPhone never compete with.
iPhone is a great phone (that you can play a little with) but a Tegra phone with Android will replace 10+ gadgets for me (GPS, Garmin Forerunner, MP3/MP4 player, media computer (720p and 1080p), web surfing computer (on my 37" TV), VOIP and yes a normal mobile phone too) and with sub $100 phones everyone will be able to own a Android phone, not just us geeks.

Android will never compete with iPhone, like that a computer never compete with a fancy gold plated calculator.

Intel Atom by itself is garbage.

On the other hand, an Intel Atom with an NVIDIA GPU is called ION.

I'd buy it if it had an ION, I do like the durability of NOKIA hardware.

Ah, does it work the same way when using dual/clone setup? And if I want to watch video fullscreen on my old TV and still can use my PC? My issue was NVIDIA dropping the FEATURE/OPTION from its latest drivers and video cards. Even Vista pulled it. :(

See:
1. http://forums.nvidia.com/index.php?showtopic=65266
2. http://forums.nvidia.com/index.php?showtopic=49636
3. http://forums.nvidia.com/index.php?showtopic=29212
4. http://forums.nvidia.com/index.php?showtopic=78251
5. http://forums.nvidia.com/index.php?showtopic=50477
6. http://forums.nvidia.com/index.php?showtopic=43756
etc.

Ah, does it work the same way when using dual/clone setup? And if I want to watch video fullscreen on my old TV and still can use my PC? My issue was NVIDIA dropping the FEATURE/OPTION from its latest drivers and video cards. Even Vista pulled it. :(

See:
1. http://forums.nvidia.com/index.php?showtopic=65266
2. http://forums.nvidia.com/index.php?showtopic=49636
3. http://forums.nvidia.com/index.php?showtopic=29212
4. http://forums.nvidia.com/index.php?showtopic=78251
5. http://forums.nvidia.com/index.php?showtopic=50477
6. http://forums.nvidia.com/index.php?showtopic=43756
etc.

Ah, does it work the same way when using dual/clone setup? And if I want to watch video fullscreen on my old TV and still can use my PC? My issue was NVIDIA dropping the FEATURE/OPTION from its latest drivers and video cards. Even Vista pulled it. :(

See:
1. http://forums.nvidia.com/index.php?showtopic=65266
2. http://forums.nvidia.com/index.php?showtopic=49636
3. http://forums.nvidia.com/index.php?showtopic=29212
4. http://forums.nvidia.com/index.php?showtopic=78251
5. http://forums.nvidia.com/index.php?showtopic=50477
6. http://forums.nvidia.com/index.php?showtopic=43756
etc.