Claim that Vista DRM causes full CPU load and global warming debunked!

Update 9/1/2007 - Gutmann paper cites shoddy web forum postings as sourceUpdate 8/30/2007 - Peter Gutmann turns to smear tactics with help from...
Written by George Ou, Contributor

Update 9/1/2007 - Gutmann paper cites shoddy web forum postings as source Update 8/30/2007 - Peter Gutmann turns to smear tactics with help from PCWorld NZ Update 8/16/2007 - Gutmann retreats and refuses to provide slides or any data to support his theories

Peter Gutmann - A security researcher at the University of Auckland New Zealand - has become one of the most cited "experts" on the evils of Vista DRM despite the fact that he never touched the Operating System at the time he wrote "A cost analysis of Vista content protection" (which he has since modified several times). He's cited by a number of folks like security researcher Bruce Schneier and has appeared on Steve Gibson's podcast raising concerns about Windows Vista DRM. There's just one little problem: Gutmann's theories are unsubstantiated and they're all wrong.

Basically the whole controversy is about the fact that Windows Vista includes HDCP (High-Bandwidth Digital Content Protection) DRM capability and the fact that HDCP includes an ICT (Image Constraint Token) flag that has the potential to slash the resolution of HD DVD or Blu-ray in half if your device isn't HDCP compliant. The reason I emphasize the word "potential" is because all of the movie studios have put a moratorium on turning on the ICT flag because HDCP market penetration is still low and there's no way they can implement ICT without generating a massive user backlash and killing any chance of HD DVD or Blu-ray adoption. I heavily criticized the ICT flag back in March of 2006 and I am on record as saying that DRM for popular music and video will die from lack of ROI, but this fear mongering on Windows Vista's inclusion of HDCP DRM is much ado about nothing. Much of it is based on hatred for Microsoft and Peter Gutmann theories are being cited as the "proof" they need.

Vista's inclusion of HDCP capability simply makes it the only Operating System that allows its users to legally play HD DVD and Blu-ray titles in addition to the fact that they can play wide-open non-DRM music and video files. It's absolutely no different from the $299 HD DVD set-top box which also includes HDCP capability. I can rip MP3s, rip DVDs, and even strip Windows Media DRM from DRM content on Windows Vista which is all contrary to all the fear mongering about Windows Vista DRM. I debunked Gutmann's scare mongering (Microsoft also responded here) back in February based on actual testing and facts but it seems Gutmann discredited theories just won't die. Not only has Gutmann not tested any of his theories or provided any data, he's now going around telling the world that Vista will cause your CPU to run at full steam which raises power consumption and causes global warming. Technology news sites like PCWorld (which got slashdotted) and the Inquirer ran with the story and took Gutmann's claims as gospel.

Quote from PCWorld: Separately, all the extra encryption required to meet Vista's content protection standards means some computer components can never enter power-saving mode, he said. Thus, when you play a movie your CPU keeps running at full steam, he said. The extra power demands make it hard to reduce electricity usage.

I emailed Gutmann and challenged him to provide data to backup his postulation and Gutmann replied:

The story is based on the slides from my talk at Usenix, with in some cases slightly selective quoting to pull out the more eye-catching claims. Once I get back I'll put the slides up on my home page and then anyone can check the details.

So Gutmann does not deny making these outlandish unsubstantiated claims and he has so far provided zero data. On the other hand, I have do have plenty of hard scientific data to refute Gutmann's claims. AnandTech ran a series of HD DVD and Blu-ray tests with hard CPU utilization numbers. On a low-end Intel E4300 Core 2 Duo CPU, CPU utilization ran as low as 7% for 1080p VC-1 encoded movie when VC-1 video compression decoding was offloaded to the ~$100 ATI 2600XT graphics adapter.

Gutmann postulated that the encryption required by Vista DRM means that it will drive CPU utilization "full steam" and he cites the fact that AES takes about 20 CPU clocks to encrypt each byte. Gutmann not only failed to test any of his theories with real-world experiments, but he didn't even bother to come up with a good postulation by doing the basic math on what 20 CPU clocks per byte means on a modern CPU. A typical 1080p HD stream is 28 mbps which is 3.5 megabytes per second which means it takes 70 million CPU clocks per second to do AES encryption on an HD stream. Since there are 1800 million CPU clocks available per second on a low-end E4300 1.8GHz dual-core CPU per core, 70 million CPU clocks constitutes 3.9% CPU utilization on one of the two CPU cores. Just doing a basic sanity check on the math before one even does any experiments will show how laughable Gutmann's postulation is. Real world testing shows that a 3.5 MB/sec HD video stream works out to be around 4% CPU utilization.

[Update 8/14/2007 - Reader "thetruth_z" says I have "limited engineering skills" saying that Gutmann is talking about encrypting the decompressed HD video which would mean 373 MB/sec (or nearly 3 gigabits per second) of payload that needs to be encrypted and decrypted. Well "truth", you might actually believe this theory of doing AES on unencrypted HD video as Gutmann may have implied, but your theory is laughable if you just do the math. Even a Core 2 Duo 2.93 GHz CPU is limited to 133 MB/sec on ScienceMark AES simulation, so it would be impossible for a high-end CPU to do 373 MB/sec. That means you would have to offload the encryption to the graphics card but we know that it's impossible to implement 3 gigabit/sec crypto offload in a $3 embedded GPU built in to a $70 motherboard. It isn't even possible on a $49 video card yet we know that a $49 ATI XT2400PRO coupled with a $100 CPU can play back full 1080p VC-1 Blu-ray video at 7% CPU utilization based on actual testing. So your theory would demand an impossible hardware feat for anything less than dedicated $500 crypto off-loader and flies in the face of common sense.

As for Gutmann's theory that this additional work load will jack up power consumption and cause global warming, it's clear Gutmann's theory on power consumption is no better than his theory on CPU utilization. In my work I do a lot of coverage of CPU performance and power consumption and I know that there isn't a linear relationship between workload and power consumption. Taking CPU clock and voltage throttling in to account, I postulated that the increase in power consumption due to a few percent extra CPU utilization will be hardly measurable.

To prove this theory, I fired up my test machine based on an Intel E6600 CPU and G965 embedded graphics and played a video in Windows Media Player classic. I noted that the total CPU utilization of around 7% (3% of that due to Task Manager) and that Intel SpeedStep caused the 2.4 GHz CPU to throttle down to its minimal clock speed of 1.6 GHz with the minimal voltage (verified by CPU-Z) and I measured 84W on the entire system. I then fired up a DVD movie at the same time with Vista's Windows Media Player and caused the CPU utilization to jump to ~15% but noticed that the CPU was still throttled to the minimum and my system power consumption stayed fixed at 84W. This proved that even an extra 8% CPU utilization makes absolutely no difference in power consumption so an extra 4% due to encryption workload means nothing.

If that isn't enough, Gutmann theorized that Windows Vista DRM and HDCP raises the cost of hardware for everyone and that even $1000 SLI dual video cards have a hard time dealing with the cost of implementing HDCP DRM. Again I ask: Where is the research? I did my research and found that a $69 AMD 690G-based integrated graphics motherboard with HDMI output has HDCP capability. That's less money than the cost of most motherboards without integrated graphics and less that the cost of some dedicated graphics cards yet it has HDCP capability. I checked with Intel and their G965, G33, upcoming G35 based motherboards that cost between $90 to $140 with integrated graphics all have HDCP capability. There are $49 video cards that implement HDCP. There are $230 22" LCD displays that implement HDCP.

Peter Gutmann if you're reading this, have you even bothered to do any research before you make your claims? As for the media that keeps citing Peter Gutmann, have you guys checked the validity of Gutmann's claims? I have thoroughly debunked Peter Gutmann's claims and it's time we put this nonsense to sleep.

[Update 8/14/2007 - Ken Fisher did a pretty good write-up here and debunked some of Gutmann's other crazy claims. Fisher noted that Gutmann had claimed (based on anecdotal evidence of what he's heard other people claim) that some people can't even properly play non-commercial HD content correctly on their PCs. Fisher noted that he has never had these problems with his HD camcorders when recording HD video to Vista. I can attest that I don't have problems with my HD camcorder and Vista either and I have never seen any proof of HDCP ICT issues with non commercial content. Peter Gutmann is beginning to look like an absolute joke with his pathetic "research" and he is reflecting very poorly on the University of Auckland New Zealand. I cannot believe that such a proud University can allow such preposterous claims from one of their staff without some sort of sanity check.]

[Update 8/14/2007 2:45PM - Ed Bott did some more debunking of Peter Gutmann here. Peter Gutmann has posted a slam at the top of his "research" paper that I didn't wait for his slides. I find it funny that he has time to write a paragraph slamming me but he doesn't have time to post the slides and he doesn't have time to post any data to support his theories. So far he's only asked others to provide data since he hasn't touched Vista yet.]

Editorial standards