I was at the Palo Alto Apple Store yesterday and saw that they had the Apple TV on display. They have three of them setup in the front right display area hooked up to some very nice Sony Bravia XBR LCD displays. The photo slideshow that was running on the unit I tried looked very nice but when I switched to watching a movie the video quality was horrible. I'm not sure what the exact resolution of the source material was but it looked no better than 480i/p and may have even been the standard 320 x 240 of iTunes videos. On top of that the bit rate the movies were encoded at was very low and there were compression artifacts all over the place. I tried three movies, The Incredibles, The Little Mermaid, and National Treasure and they all looked terrible. There was a couple standing near me at the other unit and they were commenting about how bad the video quality looked as well.
For a company that spends so much effort honing their marketing message this seems like a major blunder to me, though I don't know if this problem is unique to the Palo Alto store or endemic to all their stores worldwide. Now if these are iTunes movies (I've never used iTunes so I don't know what the videos are supposed to look like) it's commendable that Apple is following "truth in advertising" principles since their marketing slogan for the Apple TV on the Apple Web site is "If it's on iTunes, it's on your widescreen TV". However, they really need to get some DVD and HD quality movie clips on there to show off the true potential of the unit.
The other minor quibble I have with the unit I tried at the Palo Alto store is that the movie trailers that you can select don't appear to be already stored on the drive (I tried 3 of them) and you have to wait for them to download, which on the PA store's connection was really slow so I cancelled the download every time. Again it's nice that they are trying to show what the actual user experience is like but maybe they should have at least some of them preloaded as well. My guess is that some of those trailers are probably in DVD or even HD quality, given how long it was taking to download them, which would again help show off what the unit can do.
Where did you see the projected results from the Itanium? The current SPECint95 results for the latest 21264 are barely higher than the latest P3s (Coppermine) CPUs at the same clock speeds. So assuming the Itanium offers at least SOME improvement in int performance of the P3, it should surpass the 21264 on the SPECint95.
Here's some data from the latest SPECint95 results:
Compaq AlphaServer GS140 Model 6/700 700 MHz 21264A 64KB(I)+64KB(D) primary cache 8MB off chip secondary cache
SPECint95: 39.1 SPECint_base95: 34.7
Dell Precision WorkStation 410 700 MHz Pentium III 16KB(I)+16KB(D) primary cache 256KB on chip secondary cache
SPECint95: 33.7 SPECint_base95: 33.4
Comparing peak figures (SPECint95) the Alpha is 16% faster than the P3. Comparing the baseline figures (SPECint_base95) the Alpha is only 4% faster. The Alpha's primary caches are also 4 times as large as the P3s (128KB vs 32KB) and that Alpha system had a secondary cache 32 times as large as the P3s (though admittedly the Alpha's is off chip and not running at CPU speed). Both of those factors presumably have the effect of increasing the Alpha's scores. I don't know what the Itanium will have but presumably at the minimum the primary caches will be bigger than the P3s.
So basically, just from an architecture standpoint, if you factor in differences in caches sizes, the P3 is basically just as fast as the 21264 in integer performance (FP is a different story). And assuming Intel and HP don't botch the first version of the Itanium, it should offer better integer performance than the P3 (otherwise, why bother?) so at least from the artifical SPECint perspective, the Itanium should be better than the 21264.
This is really no big deal, except maybe for gargantuan products like MS Office. The Mac platform went through this same thing when switching from 680X0s to the PowerPC chips. The code for CISC-style chips is almost always more compact than code for RISC-style chips -- that's just the natural of the instruction sets.
Slashdot Mirror
I was at the Palo Alto Apple Store yesterday and saw that they had the Apple TV on display. They have three of them setup in the front right display area hooked up to some very nice Sony Bravia XBR LCD displays. The photo slideshow that was running on the unit I tried looked very nice but when I switched to watching a movie the video quality was horrible. I'm not sure what the exact resolution of the source material was but it looked no better than 480i/p and may have even been the standard 320 x 240 of iTunes videos. On top of that the bit rate the movies were encoded at was very low and there were compression artifacts all over the place. I tried three movies, The Incredibles, The Little Mermaid, and National Treasure and they all looked terrible. There was a couple standing near me at the other unit and they were commenting about how bad the video quality looked as well.
For a company that spends so much effort honing their marketing message this seems like a major blunder to me, though I don't know if this problem is unique to the Palo Alto store or endemic to all their stores worldwide. Now if these are iTunes movies (I've never used iTunes so I don't know what the videos are supposed to look like) it's commendable that Apple is following "truth in advertising" principles since their marketing slogan for the Apple TV on the Apple Web site is "If it's on iTunes, it's on your widescreen TV". However, they really need to get some DVD and HD quality movie clips on there to show off the true potential of the unit.
The other minor quibble I have with the unit I tried at the Palo Alto store is that the movie trailers that you can select don't appear to be already stored on the drive (I tried 3 of them) and you have to wait for them to download, which on the PA store's connection was really slow so I cancelled the download every time. Again it's nice that they are trying to show what the actual user experience is like but maybe they should have at least some of them preloaded as well. My guess is that some of those trailers are probably in DVD or even HD quality, given how long it was taking to download them, which would again help show off what the unit can do.
Here's some data from the latest SPECint95 results:
Compaq AlphaServer GS140 Model 6/700
700 MHz 21264A
64KB(I)+64KB(D) primary cache
8MB off chip secondary cache
SPECint95: 39.1
SPECint_base95: 34.7
Dell Precision WorkStation 410
700 MHz Pentium III
16KB(I)+16KB(D) primary cache
256KB on chip secondary cache
SPECint95: 33.7
SPECint_base95: 33.4
Comparing peak figures (SPECint95) the Alpha is 16% faster than the P3. Comparing the baseline figures (SPECint_base95) the Alpha is only 4% faster. The Alpha's primary caches are also 4 times as large as the P3s (128KB vs 32KB) and that Alpha system had a secondary cache 32 times as large as the P3s (though admittedly the Alpha's is off chip and not running at CPU speed). Both of those factors presumably have the effect of increasing the Alpha's scores. I don't know what the Itanium will have but presumably at the minimum the primary caches will be bigger than the P3s.
So basically, just from an architecture standpoint, if you factor in differences in caches sizes, the P3 is basically just as fast as the 21264 in integer performance (FP is a different story). And assuming Intel and HP don't botch the first version of the Itanium, it should offer better integer performance than the P3 (otherwise, why bother?) so at least from the artifical SPECint perspective, the Itanium should be better than the 21264.
This is really no big deal, except maybe for gargantuan products like MS Office. The Mac platform went through this same thing when switching from 680X0s to the PowerPC chips. The code for CISC-style chips is almost always more compact than code for RISC-style chips -- that's just the natural of the instruction sets.