tgdaily.com — With Intel's release of desktop processors based on its new Core Microarchitecture now expected at the end of this month, early tests from Tom's Hardware Guide are lending credence to Intel's claim that it has regained the advantage in the race for best CPU performance per watt of energy.
Jul 15, 2006 View in Crawl 4
superslothJul 16, 2006
@Maverick83: 25% performance increase is huge. Typically, next gen systems (whether CPU, GPU, RAM, HDD) only offer a 5% performance increase over last generation. This, then, is the equvalent of between 4 and 5 generations.If you can't do percentile math, thats: 105% * 105% * 105% * 105%
maskedpixelanteJul 16, 2006
If you have no preference for either brand, that's fine, go with whatever is faster today, but there's nothing wrong with somebody preferring one company over another based on a history of liking their products and service. Not being entirely brand neutral doesn't make you a fan boy. He's not saying AMD is better than Intel. He's not saying Intel's new CPU isn't better. He's clearly under no delusions here, he seems to have just had a history of buying AMD products and being very happy with them. He's not going overboard, he's not ranting, he's not even flat out refusing to buy Intel. It may just be as simple as he'd prefer to put off upgrading for the sake of buying AMD over Intel.I really don't understand why everyone's getting on his case for something as simple (and normal) as having a brand preference, especially on a site like digg where real fan-boyism is so rampant. Although, maybe it's just that people are so used to seeing fan-boys, it's just a knee-jerk reaction to anyone saying they like a company.
happyscrappyJul 16, 2006
mjPayne:First of all, the USB issue was a software issue, not a hardware one.<a class="user" href="http://www.anandtech.com/showdoc.aspx?i=2693&p=2">http://www.anandtech.com/showdoc.aspx?i=2693&p=2</a><a class="user" href="http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=16336">http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=16336</a>MS fixed it:<a class="user" href="http://www.dailytech.com/article.aspx?newsid=2417">http://www.dailytech.com/article.aspx?newsid=2417</a>So no core stepping fixed it. This is a great example of why I feel this is also a driver issue, not a chip issue.As to the caches, Intel not only has larger caches, but their caches communicate between the two cores. It gives them what AMD gets from NUMA, as long as the two cores on the same die, which they are in the 2-core case. In the 4-core case, they start to lose out again due to not having HT, but I think their speed advantage will more than carry them through.As to the 64-bit issue: you COMPLETELY mischaracterized what I said. I said Intel generally uses transistors for what the average person tends to do. And the average person doesn't do any 64-bit. I conceeded that this means Intel will be behind on 64-bit in this generation. They know it too, and likely feel that their efforts are at this time best spent elsewhere.I explicitly did not say that servers don't need 64-bit. I said this move, however wise in general, will hold them back in the server market.You mischaracterize what I said, implying I said that servers don't need 64-bit.It doesn't sound like we're too far apart here, but we are unable to debate well when you misinterpret what I say and bring in bad info.
fanonJul 17, 2006
"Yes, but it IS 4 cores of power on one board. I don't care what you prefer to call it."Be it 4 cores or 400 cores, your hardware is only as powerful as the software that utilizes it. I have a dual core CPU right now, and sure I can multitask better. But I don't have one app that utilizes both cores. Until dual core supported software is everywhere, don't expect to see much of a benefit from more than 2.
happyscrappyJul 17, 2006
Single thread performance? None of the chips we talk about are hyperthreaded. When you stall due to latency, the chip stalls, it doesn't run another thread. Now, this doesn't perhaps matter as much on servers, because servers are more I/O and transfer bound than latency-bound. They run the same code over and over, and so the caches work to reduce the latency greatly in the normal case.SUN's T1 server chip is a piece of crap. It's a niche player. It gets by because servers don't do that much CPU work. So they can get by with a very slow processor and massive hyperthreading and still get decent server performance.A good idea? Probably? But it's not going to save them. As other companies adapt good processors to use these techniques, you won't hear much about the SUN chip anymore. Because a good processor with these tweaks will outdo a less developed processor with these tweaks.As to me "conveniently leaving out the HE chips", you left out the HE chips. You mentioned the EE, I responded to your comments about the EE. I didn't mention HE because you didn't. I'm not the one claming AMD has a power/heat advantage here, so I didn't volunteer that they did.There's a 40W Woodcrest, BTW. No, it's not available today, but then again, no Woodcrest is.Performance is always part of the heat equation. If it wasn't, I'd just point at the Pentium-M ULVs. They have TDPs as low as 5W. AMD can't match that!Except performance does matter. The P-Ms are not very fast. Now AMD will have to fight the same battle. How to match Intel on performance without pulling a NetBurst and just OCing their chips until they make no sense in terms of power.
phamductriJul 20, 2006
First, I'm not an American. Just to let you guys know. Up to this point, I still don't understand why Americans would EVER drive a car like Hummer, I really don't understand. I haven't seen a car like that anywhere, not in Asia, not in Europe, only in America. It costs closed to 100$ to fill up the tank, and economically speaking, why Americans are willing to pay for that much morey for gas. Certaintly not me. Some people think it looks cool. I think it's the most unconvenient, the most idiot-looking car I have ever come across. And its mileage is single digit. Awful. Awful car