Slashdot Mirror
Core 2 Extreme 40% faster than Pentium EE 965?
Marc writes "As far as I know, this is the first time that Intel has talked about what we can expect from its new gaming CPU, Core 2 Extreme. For once, there is no word on power consumption on this new chip, but Intel talks about raw speed and a 40% gain over the current 3.73 GHz Extreme Edition 965 - which would be rather impressive and could indicate a problem for AMD. In this interview with TG Daily, Intel also claims that a Core 2 Extreme-based enthusiast PC will leave the pixel power of a Playstation 3 in the dust. Gamers, this appears to become the most exciting year for you in a long time!"
Because the ugly x86 instruction set acts as a form of compression, x86 code is more dense and fits more easily into the instruction cache than RISC code. The overhead of translating x86 to internal RISC is basically fixed and is therefore getting smaller each process shrink. It's already negligble. For this reason, the ugly x86 instruction encodings are now an advantage! x86 also gained an additional 8 registers and a cleanup with AMD64.
Hm. I knew it was one of those.
Incidentally, the 386 DX 40 was the one AMD made. Intel was rather peeved at them for licensing the design and then making it run faster than the fastest Intel chip.
Now, when you build such a high-end system you probably wouldn't skimp on the case ($200), motherboard ($200 & up), memory ($300 & up), power supply ($100 & up) and peripherals, either, so let's allow another grand for these things and you wind up with a $4000 PC.
You're about half off on that price estimate there. If you're talking about not skimping, you'd be building on a server board that's SLI capable. This means 2 processors, quite possibly 4, if Asus gets off their ass. So add on another $1k just for the extra proc (3K if it's a 4 proc board), throw in the 12-24Gb worth of high-quality registered RAM, $1,800-$2,600. Then there's cooling, you have to go liquid cooled to maintain the heat all those watts are going to put out; figure another &400-$800 worth of water blocks, pumps, hoses, reservoirs, radiators and coolant. And last but not least, we can't forget optical drives, sound card and speakers, mic, camera, media card reader and a fan controller for the fans in your radiator, figure about $600 there.
All this, and you still have to buy a monitor. Don't bother skimping on the 19", go for something with the native resolution you just paid $7-$12K to be able to handle, a 25" TFT with 8ms response time, $2500.
You think home pc's are expensive? You haven't seen anything til you make a corporate workstation meant for research, CAD/CAM or compile heave applications. I've made workstations capable of 4.86 teraflops, sucking all 1000 watts out of the wall, handling a minimum of 85 fps or so playing F.E.A.R.
"Not Skimping" are two words few people know about
Of all the Universal Constants, here's one I know: Nice guys finish last
Ok so the clock speed rocks. But does the rest of the system keep up? The big advantage I see with AMD is Hyper Transport and the newly ratified Hyper Transport 3.0. You can have a THz CPU but if you can't feed it data/instructions it's just going to waste most of it's potential.
I'm not familiar with any possible new bus technology coming out with the new Intel CPU's, but based on my current experience with the latest Dell boxes (Intel) and our new Penguin Computing and HP AMD boxes Intel has a lot of catchup to do to outperform AMD and their whole architecture.
We are using these boxes as MySQL database servers with each server containing 100+ 500 MB to 50 GB databases attached to fiber channel disk arrays. These boxes are mostly doing I/O, but a fair amount of CPU is used for sorting/math done at the database level. The AMD boxes smoke the Intel ones.
Unless Intel also releases a whole new architecture that can compete with Hyper Transport the extra speed will most likely be wasted.
This silly -Os conspiracy is starting to annoy me. -Os is actually quite a lot faster in most cases. For the total borderline cases where -O3 is faster, you're supposed to profile and change it manually. -Os has all the optimisations of O3, except for those which bloat the code unnecessarily, such as 16-byte alignment of loop headers. This type of "optimisation" bloats the code and makes it _slower_ in most cases.
The idea that Apple uses Os to make IBM look bad is totally ridiculous.
I'm so glad that AMD became a powerful player in the desktop PC and server market... not because I love AMD but because now we are really seeing some earnest competition and innovation. Before, we were happy with Moore's law, but then AMD beat Intel to 1GHz and the ensuing struggle for mind and market share has brought about some truly phenominal changes.
Keep up the excellent competition... maybe we can have a third player jump in with some new ideas? IBM? Sun? Let's see you what you have...