LOL Jordan, your are a very ebullient guy. Wow.

And don't touch TWIT? Guys, you are awesome, and deeply technical and a treat to hear, I listen to every podcast. But you aren't ready to go up against the like of Leo Laporte and friends, he is after all a podcast pioneer and has won the Podcast of the year and Person of the Year in Podcasting. How significant is he? Dvorak, with nearly 40 years of tech journalism behind him, his own pod/video cast weekly and the resources of CNET and PC magazine, makes sure he makes the show every week. Leo is that good.

However I like the thought, if you're aiming for someone, aim for the best. And I still would rather listen to Scott on CPU's than anyone else.

Excellent podcast guys! I can't wait for them to use that MotionDSP software to review all those old UFO videos. GPUs may prove or disprove the UFO footage.

Sounds like there is lots of cool stuff coming up with week.

Hey Jordan I listened to all of the TR podcast before I switched over to TWIT lol

I liked the section about AMD's roadmap.

@53min: Whoa. Perry needs to actually gee-oh-emm *read* about the OS of the OLPC. "basic" ?? It's the sexiest most exotic operating system mankind has ever devised. I know I sound a bit grandiose, but every single bit of this operating system and it's apps are not only re-programmable to the end user, but the run-time can even be viewed as it happens live. The wireless mesh network is brilliant and the range is out-of-this-world (not literally).

Anything but "basic."

Further, really? You want your kid to learn the Status Quo and forfeit his/her sense of exploration or innate curiousity and instead sell them out? I mean, I guess that makes sense...

Re AMD's Shanghai: I hope for all our sakes that AMD manages to squeeze more performance out of their K10 architecture. I know they plan to get the initial desktop processors up to 3.0 GHz (as per Chile Hardware's leaked slide), but given Nehalem's focus on multithreaded performance with Nehalem (merely 7-10% faster clock-for-clock [1] than Penryn on single-threaded workloads) I was hoping for some more general improvements from Shanghai. Having a 3x bigger L3 cache helps with that. As Damage mentions, a faster L3 and memory controller clock rate will help as well.

I hope they manage to fix whatever it was that caused K10 to stay slower than Conroe and Penryn on a clock-for-clock basis, because Intel sure isn't raising performance (through clock speeds or instructions per clock) except on loads that are already multithreaded, and they're even considering pushing back their next-generation microarchitecture after Nehalem (Sandy Bridge, that is) to give Nehalem more of a life cycle [2], so I expect there won't be much in the way of single-threaded performance improvements in the years to come from the Intel camp. I for one value single-threaded performance a whole lot more than multi-threaded performance, and I agree completely with Damage in that increasing per-core performance is essential.

[1] http://www.google.com/search?q=cache:www.hexus.net/content/item.php... (the original seems to have been pulled)
[2] http://en.expreview.com/2008/09/04/lynnfield-has-powered-on-and-boo...

Re Fiorano: Intel has VT for Directed I/O (VT-d) which is their IOMMU. [3] It's existed since the 5400 series chipsets for Xeon and since the Q35, X38 and X48 chipsets for the desktop. It's about time AMD added IOMMU support in. (I think it's called AMD IOV technology, at least as per a presentation from AMD [4] regarding their virtualization improvements.) There's apparently a way to fake an IOMMU using the GART on K8-generation processors and above [5], but it doesn't seem to be conducive to e.g. assign devices to virtual machines the way Intel VT-d and AMD IOV enable [6], so the VMs can use native drivers to control the device and cut the hypervisor out of the I/O flow of control to allow for improved peformance.

[3] http://www.xen.org/files/xensummit_4/VT_roadmap_d_Nakajima.pdf
[4] http://xen.org/files/xensummit_fall07/11_TomWoller.pdf
[5] http://www.nabble.com/-PATCH--Enable-K8-GART-as-an-IOMMU-td8929114....
[6] http://www.nabble.com/vPro-(VT-d)-support--td12400047.html

Re Magny-Cours: pronounced "many cores", it's a pun.

Re Penryn speed bumps: oh really? *strokes nonexistent beard* Dang. I'd assumed that Q9650 and E8600 were the highest that Penryn was going to go, at least on the desktop.

Re Bulldozer: That's too far out (2011 [7]) I suppose.

[7] http://news.cnet.com/8301-13579_3-9938372-37.html?tag=nefd.lede

Re Aspire One: Good to hear Perry Longinotti on the podcast. I think the Aspire One is the one notebook with the best constellation of price, processor performance, battery performance and keyboard arrangement among all the netbooks:

* the Dell Inspiron Mini 9 and Lenovo IdeaPad S10 have the right price but a messed-up keyboard.
* the HP Mininote 2133 has the right keyboard and battery performance, but a price that's way too high and processor performance that's easily the lowest.
* the Asus Eee PC 900, 901, 904, 1000, 1000H, 1000D and 1000HD have good processor and battery performance (not least because many of these models come with SSDs and they all have 6-cell batteries, standard) but the 8.9" models' keyboards are painful to use and they're also overpriced.
* and the MSI Wind U100 has a good keyboard and good processor performance, but comes with the 3-cell battery as standard and the $100 price hike made it a rather unattractive option.

That essentially leaves the Aspire One as the best of all worlds, as long as you pick up a model that comes with a 6-cell battery. It's cheap, it comes with an Atom N270 rather than, say, a VIA C7, it is offered with a 6-cell battery on some models [8], and it has one of the larger keyboards among the 8.9" netbooks. (And, hey, it has a LED-backlit screen.)

I'm actually thinking that a Vostro 1310 off Dell Outlet [9] is an excellent choice if the tiny form factor and weight of an ultraportable isn't a high priority. It's priced above the cheap-netbook range, but for as much money as an Eee PC 1000 or an HP Mininote 2133, you get a whole lot more laptop. And it's built like a business-grade notebook, complete with magnesium alloy frame, unlike the cheap Inspirons, and it isn't exactly huge, at 13.3".

Perry's absolutely right in that the more expensive netbooks are pretty pointless. Fujitsu-Siemens for instance is renowned for their mainstream ultraportables in the LifeBook P range, and would be preferable over netbooks that approach that level of price. So are the Lenovo X series notebooks, the Dell Latitude D4xx and E4xxx, the Toshiba Portege series and the HP 2xxxp series notebooks.

The Classmate PC is actually fairly successful out here in the "third world". For instance, here in India, HCL makes MiLeap X Classmate PCs [10]. I'm not sure it's actually being used in schools, though; I think people have bought them primarily as cheap computers, not necessarily specifically for kids. I think they have a pretty decent-sized market among college undergraduates, though.

[8] http://search.ncix.com/displayproductdetail.php?sku=31428&vpn=LU.S0...
[9] United States: http://outlet.us.dell.com/ARBOnlineSales/topics/global.aspx/arb/onl... or United Kingdom: http://outlet.dell.com/Emea_Dfo/EuDispatcher?target=InventoryPage&a...
[10] http://www.pcworld.com/businesscenter/article/141445/intels_classma...

Re SoftKinetic: Speaking of camera tracking, a Carnegie Mellon researcher has written software that allows head tracking, finger tracking (for a virtual whiteboard), and so on using a stationary Wii remote and visualizes it. (Complete with cool videos.) [11] It's not as sophisticated as what SoftKinetic seems to do, but it's pretty neat for something as inexpensive as the Wii remote.

[11] http://www.cs.cmu.edu/~johnny/projects/wii/

Re MotionDSP: Maybe we'll have the CSI-style ENHANCE! ENHANCE! ENHANCE! become reality. :p

Re Badaboom Media Converter: It's too bad it's so picky. I hope that x264 or similar project will end up creating an OpenGL GLsl shader based encoder (or failing this, CUDA-based). I get the feeling it'll end up being at least as good as the Badaboom Media Converter, and hopefully better.

Re Google Chrome: I doubt I'll end up opting for it. I use Opera as a feed reader, and its image management is essential for me (specifically being able to run it with image loading turned off and being able to load images rapidly with a single keystroke), as is its native gesture support. Also, the fact that it doesn't have separate processes for each tab means that it doesn't use unholy amounts of RAM. (Chrome is pretty bloated [12] although it can at least be made to run with plugins in-process [13].)

Apparently nightlies of Firefox 3.1 (but not the recently-released alpha) have a new JavaScript engine called TraceMonkey [14] that sounds like it works a lot like V8 and it also runs faster than V8 [15] (although it's still in development).

Opera has had the tabs-on-top and tab-private-addressbar organization for a long, long time, by the way [16].

[12] http://exo-blog.blogspot.com/2008/09/google-chrome-fattest-of-them-... and its follow-up post http://exo-blog.blogspot.com/2008/09/ie-8chrome-follow-up.html -- at least it's slimmer than IE8 beta 2.
[13] http://www.myscienceisbetter.info/2008/09/install-google-chrome-on-...
[14] http://arstechnica.com/news.ars/post/20080822-firefox-to-get-massiv...
[15] http://arstechnica.com/journals/linux.ars/2008/09/03/new-firefox-ja...
[16] http://www.opera.com/products/desktop/screenshots/

Great episode TR guys!! Really enjoyed.

Jordan you are the real driver, you delivered my questions precisely the way I wanted, and thanks Scott for your views and data directly from AMD.

Another very special part I like is the MotionDSP. Very well done.