SlowSpyder
Lifer
- Jan 12, 2005
- 17,305
- 1,002
- 126
Originally posted by: JackyP
True, but I'm a sceptic, though. A follower of the scientific method.Originally posted by: SlowSpyder
Originally posted by: JackyP
Somewhat comparable? Yes 80% is somewhat comparable to 100%, still it was much slower - end of story.Originally posted by: Cogman
Originally posted by: Stoneburner
Unspossibles! THis may match i7 at 5.3 ghz!
I hate when these kinds of numbers drop before the important stuff.
Phenom was somewhat comparable to the C2Q, the biggest holdback for it was the the C2Q was able to overclock like mad. With the possibility that the phenom II overclocks like crazy, could be comparable to Penrynn in speed, and will possibly (most likely) be cheaper then intels offerings is good news.
No one cares about overclocking in terms of market size and money, so that surely wasn't the biggest holdback, stock clock, powerconsumption and performance was, but it's still lovely to see such a nice clocking chip from AMD.
Shanghai may be cheaper to the end user, but at the cost of margins. Shanghai's die is still at least 20% bigger than penryn, monolithic and using a new process. So the chip per se is not cheaper (i.e. cheaper to produce) at all.
I really think Phenom was more competitive then you give it credit for. The biggest problem with Phenom in my eyes was the performance you got for the amount of power it consumed. Intel really hit it out of the park with the C2D, it was almost always faster then Phenom, and did so using less power.
But I do, that's why I mentioned it. 243mm² (263mm² including test logic, don't know if this makes it into the final product). Penryn 2*107mm².I don't know the die sizes off the top of my head, but my guess is a 45nm Denab isn't much bigger in size then two Penryns added togeter for the Intel quad.
I think that the Intel "glue approach" has some advantages and drawbacks when it comes to yields (one error does not kill your 214mm² chip, just a 107mm² one, but the glue is not free either).
I'm not considering price for the end-user, but for the company. They could give chips away for free if they wanted to, still below certain margins they cannot survive long term, that's why I mention die size.Also, I think AMD has shown that they can have a pretty good product that isn't the absolute fastest but still fast enough so long as it's priced right. Take a look at the Radeon 4850/4870.
I was talking about penryn, thank you for proving my point. There are hundreds of reviews out there, I think -20% for barcelona is even pretty generous. So much for beating a dead horse, I just don't like lies...80% Try 85-90% at the same clock speed, with a lower price (per cpu and whole system). I just ran the numbers, 77% was the absolute lowest it did, and it quite a few cases it was 95% of the speed of a Q6600 at the same clock speed (the 9700 that is).
Yes its slower, however you are paying less for it and the entire system. Does it cost AMD more to produce a CPU, probably, But I wasn't speaking about their benefits as the producer.
Actually I have been talking about the "benefits to the producer". If AMD is broke you won't get anymore chips from them, remember this very well.
Tough times are ahead, let's hope this is enough to keep them afloat.
Ok, so Intel does have a smallish advantage in die size, but those numbers look fairly comparable. The current 65nm Phenom is huge compared to 2 Penryn cores, this puts them in the same ball park. I think the Intel 'glue' approach has an advantage or two as well, I'm sure it's a lot easier for Intel to have made their quad core then AMD. Intel designed a dual core and was able to just put two on a single package for a quad, I would imagine that shaves off a lot of time in getting the part to market compared to the 'native' quad core approach AMD went with. Also, a bad core doesn't kill AMD's chip either, they sell it as an X3 part. On the X3 parts I don't know if AMD will make money, break even, or sell at a loss, but any of those options are better then just tossing a chip with a single bad core in the garbage.