Maximum PC CUDA RSS Feed

Nvidia GPU Technology Conference Liveblog

Will Smith — Wed, 30 Sep 2009 17:07:56 -0500

Nvidia GPU Technology Conference Keynote Liveblog

Dell "Personal Supercomputers" Now Available With Nvidia Tesla GPUs

Andy Salisbury — Thu, 07 May 2009 16:27:48 -0500

To those looking for another venue to get their very own supercomputer, you’re in luck! Nvidia has recently announced that their CUDA-based Tesla C1060 GPU is available in Dell’s Precision R5400, T5500 and T7500 workstations effective immediately.

If you’re worried that just one of these GPUs isn’t enough to handle your hardcore needs, worry not – just one C1060 has enough power to control the main system of the European Extremely Large Telescope project (reportedly the world’s largest). According to Jeff Meisel with National Instruments, a workstation “equipped with a single Tesla C1060 can achieve near real-time control of the mirror simulation and controller, which before wouldn't be possible in a single machine without the computational density offered by GPUs."

Image Credit: Nvidia

Sparkle Introduces 2GB, Single-GPU Graphics Card

Pulkit Chandna — Wed, 08 Apr 2009 20:21:12 -0500

Taiwanese VGA card manufacturer Sparkle has introduced what it claims is the world’s most powerful single-GPU graphics card. According to Sparkle, its GeForce GTX 285 is the only single-GPU graphics card to feature up to 2GB GDDR3 memory. However, on closer inspection of its claim, we found another company with a similar offering.

Palit Microsystems began offering a custom-built GTX 285 with 2GB memory in February. From the face of it, Sparkle’s entire staff was probably marooned on a remote island – or away on an intergalactic excursion, and therefore had no idea what was going around.

The GTX 285 runs at a core clock frequency of 648MHz. Sparkle has also promised its card will deliver “30% faster performance than competing single GPU graphic card solutions.” But the company is mum on pricing.

Image Credit: Sparkle

Nvidia: It’s Time to Kill CUDA

Will Smith — Mon, 12 Jan 2009 14:22:02 -0600

Like many of you, the first real 3D accelerator I owned was a 3dfx Voodoo card. This was way back in 1995. DirectX and Direct3D had yet to be released to the public, and OpenGL was only used for CAD and scientific rendering apps. In those primordial times, if a game developer wanted to harness the awesome rendering power of the Voodoo hardware, he had to write his game with Glide, 3dfx’s own application programming interface (API). This was all before the open standards movement became a powerful force in development circles, and Glide offered 3dfx a major competitive advantage: If a gamer wanted to see all the kick-ass 3D effects that Glide enabled, he had to play the game on 3dfx hardware—lest he suffer Glideless, in a depressing, busted-up world of jaggy, unfiltered textures.

The 3dfx/Glide domination ended when id Software and other game developers started releasing titles that used the OpenGL API, which wasn’t dependent on 3dfx hardware (but worked with 3dfx chips through a Glide translation layer). OpenGL opened the door for other 3D chip companies to build competitive products, and thus ATI, S3, Matrox, and Nvidia entered the fray with hardware of their own.

With every new OpenGL or DirectX game released, Glide slowly transitioned from an advantage to a liability for 3dfx. As competitors like Nvidia embraced new technology and embarked on a period of incredibly rapid improvements, 3dfx remained tied to its Glide past, and, as a result, was slow to embrace new rendering enhancements, such as 32-bit color and antialiasing. Ultimately, this contributed to 3dfx’s demise, and embracing open standards allowed Nvidia and ATI to flourish.

Why are we talking about this today? Because Nvidia stands at a crossroads, with two closed, proprietary APIs that have mainstream potential: the general-purpose computing CUDA API, and the PhysX physics-acceleration API, which sits on top of CUDA. These are both promising technologies, but only owners of Nvidia hardware can harness their power. Meanwhile, there are two emerging open standards that mirror what Nvidia is doing with its proprietary development. One is OpenCL 1.0, and the other is a general-purpose GPU computing API, which Microsoft will include in DirectX 11. There are a relatively small number of consumer applications that use CUDA, PhysX, or OpenCL right now, but the possible applications for the tech are endless—grossly simplified, these APIs let graphics chips perform CPU-like functions. The question Nvidia needs to be asking is simple: Will developers write their general-purpose GPU computing apps using a proprietary API that works on only a subset of PCs—those stuffed with Nvidia hardware—or will they use an open API that will work on every PC on the market?

Nvidia’s path is clear: It needs to stop trying to convince us that closed APIs are good, and instead embrace OpenCL and Microsoft’s yet-to-be-named solution. It needs to port PhysX to run on one of the open APIs, then use PhysX as a platform to advertise the kind of power that Nvidia delivers (with the recipients of all this messaging being ATI diehards and anyone considering the forthcoming Larrabee GPU from Intel).

By focusing on what its always done well—building kick-ass hardware—instead of force-feeding us closed APIs, Nvidia will thrive. As for CUDA? It’s served its purpose, but its time has passed. It’s time to kill CUDA.

No BS Podcast #89: Gordon Brings the Rage

The Maximum PC Staff — Fri, 12 Dec 2008 12:52:08 -0600

This week, we recorded a mostly zombie-free edition of the No BS podcast. While there was a little undead chat, we also talked about CUDA vs. OpenCL vs. DirectX 11 and using iTunes the Gordon Mah Ung way. This week, we're pretty certain that we even managed to post the right pocast (if you missed last week's, just redownload it. It's linking to the right one now). Join the podcast gang as we answer your tech questions, take a trip to the Lab, and get a chock-full-o'-rage edition of Gordon Mah Ung's Rant of the Week!

Do you have a tech question? A comment? A tale of technological triumph? Just need to get something off your chest? A secret to share? Email us at maximumpcpodcast@gmail.com or call our 24-hour No BS Podcast hotline at 877.404.1337 x1337--operators are standing by. For the love of all that's holy people, if you guys don't start asking tech questions, we're going to change the name to the Nothing But Undead podcast...

Subscribe: http://feeds.feedburner.com/maximumpc/1337

Russian Firm Breaches WiFi Security Bulwark Using Nvidia Graphics Cards

Pulkit Chandna — Fri, 10 Oct 2008 18:46:20 -0500

Elcomsoft, a Russian firm, has devised an ingenious method of breaching WiFi’s WPA and WPA 2 encryption systems, according to a report at SC Magazine.

The technique leverages the parallel processing power of Nvidia’s latest graphics cards to speed up the “password recovery” process by 10,000 per cent. Global Security Systems (GSS) has advised enterprises to deploy VPNs for safeguarding their WiFi networks.

We, too, can only advise you to secure your office WiFi network using VPN encryption before professional industrial sleuths start waging brute forcing blitzkriegs using ordinary graphics cards.

GPU-Based Video Encoding is Here: Does That Mean Curtains For Your CPU?

Will Smith — Wed, 10 Sep 2008 11:01:00 -0500

Earlier this summer, both Nvidia and ATI hosted press events to unveil their new hardware—and the excitement about GPU-based encoding was palpable. We were promised that our videocards would make Photoshop faster and better and our GPUs would encode video 10 times faster than our CPUs. In fact, someone lacking tech savvy would have left these presentations thinking, "Wow, these GPU things can make common computing tasks run insanely fast, and there are a couple of games that work with them too." Of course, as is typical, the truly big promises (like 10x faster video encodes) were off in the future, when the software was "ready."  

Well, the software's nearly ready. Elemental's Badaboom uses Nvidia's CUDA interface to do lots of the grunt work of DVD ripping by using the GPU instead of your musty old CPU. I've been in the Lab for the last few days putting this app through the ringer. Our test bed for this challenge is an Intel Q6600 quad core, running at a stock 2.4GHz, with 4GB of memory and a GeForce GTX 280 reference board.  

The short, short version, for anyone who wants to skip ahead, is that Badaboom is an extremely promising first-generation utility, and there's a lot to like. However, there are some serious problems for anyone looking to replace Handbrake (www.handbrake.fr) or AutoMKV (http://forum.doom9.org/showthread.php?t=134478) for their archival DVD rips.   

To properly evaluate Badaboom, we compared its speed and video quality to Handbrake’s, which is essentially a GUI front end for a tweaked version of the CPU-powered X264 encoder. Unfortunately, we quickly discovered a hitch: Badaboom isn't capable of running H.264’s maximum quality settings, as Handbrake is. By default, Handbrake uses the H.264 high profile, while Badaboom uses the lower-quality baseline profile. To see how both apps performed in an apples-to-apples scenario, we ripped a copy of Pee-Wee's Big Adventure to our hard drive and then ripped it to H.264 with each app, using a variety of settings. Here's our test matrix:

Settings for Test Runs

	Elemental Badaboom Beta 2	Handbrake 0.9.2
Test #1: PS3 Stock Settings	upscales to 1280x720 (?!), 2500 kbits/sec, default settings	PS3 Stock settings, Single-pass, 720x480 resolution
Baseline Settings	PS3 preset, 2100kbit/s	2100kbit/sec, single pass, Handbrake using baseline settings(see end of article)
"Best" Quality	Custom Media Center Preset - 2500kbit/s	Deux Six Quatre preset - 2500kbit/s
iPhone	1500 kbit/s, preset settings	1500kbit/s, preset settings

PS3 Stock Settings

When using two different codecs, even when you're encoding the same content, it's extremely difficult to ensure that you're actually measuring similar workloads. As a case in point, examine the first test we ran, the PS3 stock settings. We chose the PS3 profile because video ripped for the PS3 will also work on many other devices, and includes options that are commonly used while ripping.

While Handbrake ripped the DVD video at native resolution using the PS3 preset, the default Badaboom PS3 profile ripped the disc using the H.264 baseline preset and then upscaled it to 1280x720. The two apps showed remarkably similar results, separated by about 3 minutes on a 40-minute rip, but the quality of the Handbrake encode was vastly superior, despite the higher resolution of Badaboom's output. When we ran the test again, without forcing Badaboom to upscale the video, our results strongly favored Badaboom. However, since Handbrake runs a High profile H.264 encode by default, we wanted to see how the GPU stacked up against the CPU when the workloads were more equal. (Be sure to click through to see the native-resolution, losslessly-compressed image. There are additional quality comparison images on the final page of the story.)

PS3 Stock Settings: When we stopped Badaboom from upscaling the video to 720p from DVD resolution, we saw a substantial speed boost for the GPU-accelerated app. However, Handbrake was encoding using the much more strenuous High H.264 profile. (click image for full-size)

TEST #1: PS3 STOCK SETTINGS

	Elemental BadaBoom (PS3 preset, 2500bits/sec, 720p)	Elemental BadaBoom (baseline, PS3 preset, 2500bits/sec, 480p)	Handbrake .9.2 PS3 (high profile, 2500bits/sec, 480p)
PeeWeeMark (sec)	2256	1097	2460
File Size (gb)	1.56	1.53	1.7

Test #2: Baseline Settings

That's when we went to the video codec experts at Doom9.org and asked how to configure Handbrake to do a fair encoding test, in order to most effectively mimic the settings that Badaboom uses in Handbrake's X264 codec. After a few misfires, we settled on a fair baseline that uses roughly the same features and results in roughly the same image quality as Badaboom (The full settings are listed at the end of the article if you want to see the specific X264 options we used). We were surprised by the results. In a direct comparison, using as close to the same visual quality settings as we could, Handbrake's circa February 2008 X264 codec actually beat the Elemental encoder by almost a minute. Image quality was roughly the same; we've included several stills below so you can directly compare the results.

Baseline comparison: In this relatively high-motion scene, you can see that both codecs look roughly the same. Both suffer some compression artifcats, but as a whole the quality is roughly equal.

TEST #2: BASELINE "APPLES-TO-APPLES"

	Handbrake custom baseline (see command line at end of article)	BadaBoom PS3 (480p)
PeeWeeMark (sec)	1059	1097
File Size (gb)	1.43	1.7

Test #3: Best Quality Settings

Next up, we tested using the highest quality presets in both Badaboom and Handbrake. If the PS3 setting used above is a format commonly used by folks who want to stream video to their game consoles, this is the format for people who want true archival quality backups of their DVDs. The Handbrake Deuce Six Quatre profile uses a two-pass encode to transcode your video to a video that's nigh indistinguishable from the original DVD at a fraction of its original size. On the other hand, Badaboom uses the exact same baseline profile, scales the video up to 720p, and sets the average bitrate to a whopping 8,000kbit/sec. The end result for Handbrake was an encode that took about three times longer than Badaboom, but it created a file that's half the size of the and looks twice as good as the GPU-accelerated transcoder.

Best Quality: At the best quality settings for the two apps, Handbrake took much longer to complete, but the final encode was approximately half the size, and was the best looking of all of our test videos. The compression errors that we saw in previous test runs were exacerbated here by the unnecessary upscaling to 720p.

TEST#3: BEST QUALITY

	Badaboom Custom Media Center Profile	Handbrake Deuce Six Quatre
PeeWeeMark (sec)	2472	7142
File Size (gb)	4.58	1.84

TEST #4: THE IPHONE CONVERSION

Things got a little more interesting for Badaboom when we started looking at decreasing the resolution at the same time we transcoded from MPEG2 content to H.264—as you would when you convert a movie to play on your iPhone. Performance damn near kicked ass for this test run, and we don't really mind sacrificing a little quality for a much faster transcode, especially since we'll be watching these videos on a tiny phone screen. So, how much faster is Badaboom in this test? The GPU-accelerated app finished the video transcode in approximately half the time it took Handbrake (again, using the presets, but using the same 1500kbit/s bitrate). Check the stills below for the image quality differences.

The iPhone Test: This is where Badaboom really shines. When scaling your video down to a lower resolution at the same time it transcodes, Badaboom absolutely crushes Handbrake, and because it's for viewing on a pocket-sized screen, we barely notice the image quality hit.

TEST #4: ENCODING FOR IPHONE

	Badaboom Custom iPhone	Handbrake iPhone
PeeWeeMark (sec)	728	1548
File Size (gb)	0.074	1.04

The Upshot

I was surprised by Badaboom—both by the image quality it produces using the baseline H.264 profile, and by its performance. However, I'm not surprised that the nascent app didn't deliver a 10X H.264 video encoding perforamnce boost, despite the benchmarks Nvidia demoed this spring and summer that seemed to indicate GPU-based encoding was insanely fast. When compared in an apples-to-apples fashion with a multi-threaded encoder running on the cheapest quad-core CPU available (2.4GHz Q6600s are available at Newegg for $185 today), the results are much less impressive than the 8x to 10x speed up promised by Nvidia's demos (which used a single-threaded iTunes plugin running on the same slow quad core, so only 25 percent of the rig's CPU power was being used). 

For video encodes, Badaboom is promising. It will be more so when Elemental decides to support the H.264 high profile, which it must do to compete with free CPU-based applications. For people who want to archive their DVD collections, the baseline profile is insufficient. However, Badaboom is extremely compelling for converting videos to play on portable devices. The time savings for these transcodes are huge, whether they’re from an already converted file to a lower resolution and different bitrate or straight from the DVD source. More importantly, the image-quality sacrifices Badaboom imposes are less of a concern when watching movies on your phone's screen. While we wouldn't recommend Badaboom for people who want to make DVD backups for playback on the big screen, if you simply want a quick way to get video on your iPhone or PSP, it might be worth the $30.

I did experience a few glitches during my testing using Badaboom. These glitches manifested in different ways, but they typically affected the entire screen for just a few seconds. My personal favorite glitch made the Warner Brothers logo at the start of the movie we tested with look like a backdrop from the beautiful Xbox game, Braid. Check it out below if you want to see more.

Codec Concerns: Despite delivering decent performance, we found multiple encoding glitches in almost every video we created with Badaboom.

Am I excited about the promise of GPU-based computing? Definitely. Am I going to spend $30 on Badaboom when it launches? Probably not, at least not without better-quality H.264 encoding. However, despite a few first-gen problems, Badaboom is off to a good start. Frankly, I don't understand how Elemental can reserve high-profile support for the more expensive Pro version when several free apps support the highest quality settings H.264 allows. The good news for Badaboom is that performance is only going to improve as the software matures and the hardware improves.

Does GPU-based computing deliver enough for us to trade in our quad cores for "optimized" PCs today? Hell no. While no one at Maximum PC thinks those fancy $1,000 CPUs that Intel sells make much sense for a value-oriented system builder, there simply aren't enough non-gaming applications that take advantage of the GPU today for us to consider dropping from a quad core to a dual core when quad-core CPUs are available for under $200. A quad-core CPU and a good codec deliver similar performance and superior image quality compared to the fledgling GPU-based encoder, and a quad-core CPU speeds more than just video encoding. Whether you're running four single-threaded apps or a single multi-threaded app, the CPU is too flexible to consider stepping back to a dual or single core to save a mere $15. The way we look at it, $200 quad cores mean that even gamers on a tight budget should be able to afford a kick-ass videocard and a kick-ass CPU.

Handbrake lines: 

Handbrake Baseline: -i "C:\DVD Rips\PEE_WEES_BIG_ADVENTURE" -o "E:\Pee-Wee Handbrake Baseline.mp4" -e x264 -E faac -p -b 2100 -x subme=1:partitions=i4x4:keyint=120:aq-strength=0:me=dia:vbv-maxrate=14000:no-cabac:vbf-bufsize=2100:no-dct-decimate:level=3.1:no-b-adapt:scenecut=1:progress 
PS3: -i "C:\DVD Rips\PEE_WEES_BIG_ADVENTURE" -o "E:\Pee-Wee Handbrake Baseline.mp4" -e x264 -E faac -p -b 2500 -x level=41:subme=5:me=umh -B 160 -R 48 -D 1 -v 
iPhone: -i "C:\DVD Rips\PEE_WEES_BIG_ADVENTURE" -o "E:\Pee-Wee Handbrake Baseline.m4v" -e x264 -E faac -w 480 -b 960 -I -x level=30:cabac=0:ref=1:analyse=all:me=umh:subme=6:no-fast-pskip=1:trellis=1 -B 128 -R 48 -D 1 --markers="C:\Users\Will\AppData\Local\Temp\chapters.csv" -v
 Deuce Six Quatre: -i "C:\DVD Rips\PEE_WEES_BIG_ADVENTURE" -o "E:\Pee-Wee Handbrake Baseline.mkv" -e x264 -E ac3 -p -b 1600 -2 -T -x ref=5:mixed-refs:bframes=3:bime:weightb:b-rdo:b-pyramid:me=umh:subme=7:trellis=1:analyse=all:8x8dct:no-fast-pskip -D 1 --markers="C:\Users\Will\AppData\Local\Temp\chapters.csv" -v

More Images

Baseline

Baseline comparison: In this relatively high-motion scene, you can see that both codecs look roughly the same. Both suffer some compression artifcats, but as a whole the quality is roughly equal.

Baseline Still Comparison: Again, both encoders produce roughly the same quality level. Check Pee-Wee's jacket, the skin on his cheeks, and the vertical lines on the Wig Factory's storefront to see potential compression problems.

Baseline Still #2: Pay extra close attention to the sweat on Pee-Wee's brow. I

Best Quality Comparisons

Best Quality Motion: In this instance, upscaling the resolution and wildly increasing the bitrate don't really show any improvement on general visual quality. The edges of Pee-Wee's shadow are still jaggy, and the scene under delivers. On the other hand, the Handbrake Best Quality rip looks awesome.

iPhone Comparisons

SIGGRAPH 08: Nvidia Brings Along D Series Qaudro Plex Visual Supercomputers

Pulkit Chandna — Fri, 15 Aug 2008 09:42:05 -0500

Two years ago, Nvidia unveiled its Quadro Plex range of visual computing systems at SIGGRAPH 2006. Now, at this year’s SIGGRPAH, it has announced desk-mounted visual supercomputers in the Quadro Plex range. The D series of Quadro Plex visual computer systems is claimed to have leapfrogged previous versions by over a 100% in terms of performance. The NVIDIA Quadro Plex 2200 D2 VCS has two Quadro FX 5800 GPUs, 4 dual-link DVI channels, and 8 GB of frame buffer memory. Whereas its sibling the NVIDIA Quadro Plex 2100 D4 VCS has four GPUs, 8 dual-link DVI channels and a 4 GB frame buffer.

The D series visual supercomputers are ideal for highly taxing 3D models, engineering designs and other scientific visualizations. The hundred of Nvidia CUDA Parallel Processing Cores pack copious parallel computing capabilities and the visual supercomputers can be easily hooked to workstations or servers using PCI Express adapter cards. The D series is due in September with prices starting at $10,750.