TPU ChatChat on IRC
Internet Relay Chat?
Java Applet client
Who's on IRC?
Posted by FatalDragon [send private reply] at June 10, 2002, 09:18:05 PM What is TRUELY the fastest processor? Is the Xeon and UltraSPARC III faster than they are advertised? Like is a 1.0GHz UltraSparc faster than a 2.0GHz P4, since UltraSparc has 2MB cache? Are there any other CPUs other than Intels, AMDs, Sun, and Cyrix? ThanX!
Posted by CodeRed [send private reply] at June 11, 2002, 03:34:25 AM My 1.0 Athlon is faster than a 2.0 P4. Before everyone yells bullshit let me explain. See, Intel enjoys screwing its customers up the ass. The P4 is the biggest piece of shit to ever be released. I could go on and on about it's problems but this guy does a much better job:
Now, here is where the ass rape comes in. Intel rushed the release of this piece of garbage to make quick cash (AMD was making them sweat for a while) and as a result they released a "crippled" processor. Several months after the release of the P4 came the P4A (Known as the Northwood core). This is what the P4 was supposed to be, and it is a decent processor (Still underpowered compared to equally clocked AMD's) I know it sounds like I am an Intel basher, but I do give credit where credit is due, the P4A has many attractive features. First of all they run VERY cool (as in temperature). Many people have overclocked their 1.6A's to 2.0 or higher with the stock HSF. The 2.0's have been pushed to 2.8 with stock cooling. On top of that their L1 cache is gigantic and they support DDR and RDRAM memory. Now comes the cons. They are very expensive. They come with an integrated heat cap (If you OC you know why this is a con). And like I said earlier, an equally clocked Athlon still blows them away (MHZ's are meaningless, don't buy into the hype).
AMD has been showing off it's new Clawhammer chip for about a month now. I'll admit I don't know much about it, but I did see a couple benchmark results, and according to SiSoft Sandra 2001 an 800mhz Clawhammer outperforms a 2000mhz P4A by about 20%. This is obviously very exciting, as these chips are expected to reach 3-4ghz. A little farther down the road (1st quarter '03) the Clawhammers big brother will be released, the Sledgehammer. This will be a 64bit processor manufactured using .13 micron technology. It will have an on-chip memory controller and will support hundreds of gigs of memory. This chip is expected to break the 5ghz barrier.
As for right now, honestly, Intel has the fastest chip. The 2.0 P4A, when overclocked, will outperform the best AMD on the market (the Athlon XP2100). But in the long run my money is on AMD. The Sledgehammer will annihilate Intel's Katmai and I can't wait to see it happen.
Posted by metamorphic [send private reply] at June 11, 2002, 05:00:05 AM Also, clock speeds are not all that matters. Intel only do well with P$'s becasue they advertise them at their max clock speed which is now around 2 GHz. AMD however realize that other things are important like L1 and L2 chache, onboard memory, bus speed and many other things, not just the amount of calculations it can do. If you have a P4 2.4GHz, it will outpreform it the current best athlon (2200+XP) by some way at pure number crunching apps and generating things like fractals. However at day to day things such as databases, spreedsheets and many games the althon does just as well. The athlon 2200+XP however is only clocked at 1.67GHz compared to intels 2.4 GHz P4. Yet it still does almost as well and costs half the price.
Posted by unknown_lamer [send private reply] at June 11, 2002, 10:07:05 AM The P4 is not a crappy processor...applications just haven't been optimized for it yet :).
The same thing happened with the i686 (PPro) was released.
The UltraSPARC should be faster than a P4 because it is a 64-bit chip
and can process twice as much data per instruction (or something like that,
it all depends on your app. E.g. a program adding large number would run
much faster on a usparc because it could use 128-bit long long numbers and
be able to go much longer before it had to revert to software bignums).
I can't wait for the AMD x86-64. Debian already has a port ready and GCC 3.1
can target it (32 and 64 bit code even with -m32 or -m64). I think a dual x86-64
machine would be quite fast...(especially with 8x AGP and the new Matrox card!)
Posted by FatalDragon [send private reply] at June 11, 2002, 10:10:46 AM Isnt the L1 and L2 cache how much data can be processed at a time? Is the UltraSparc III faster than P4 even though its MHz arent as high? Is the Xeon faster since it has a 2MB cache? ThanX!
Posted by FatalDragon [send private reply] at June 11, 2002, 10:12:28 AM Isnt the Sledgehammer now the Operton?
Posted by unknown_lamer [send private reply] at June 11, 2002, 10:24:26 AM The Sledgehammer is the Operton now (or was that the clawhammer that was the x86-64?)
The amount of L1 and L2 cache are very important, but they don't determine how much data you can process at once. They are exactly what their name says they are--cache. The real amount of data you can process at once is what is in your registers. Having a lot of cache makes spilling the registers much faster, but that is still no substitute for not having a lot of registers. The truth is that the x86 architecture is really, really register poor. The UltraSparc has a lot of registers, whereas the p4 doesn't have nearly as many (I can't find exact numbers, but let's just say that the p4 is crying for registers). The Usparc also has larger registers...the end result is that the usparc will spill its registers less often giving you better performance because it doesn't have to take the hit from fetching from cache or ram as much (which is a very expensive operation). The Moral of the Story: mo' registers is mo' betta (hah! I'm a homey now)
Posted by CodeRed [send private reply] at June 11, 2002, 01:07:21 PM "Isnt the Sledgehammer now the Operton?"
Yes, I know, I don't know why though, Sledgehammer sounds a lot cooler. If you want pics check out my thread called "My next system"
Posted by Mycroft [send private reply] at June 11, 2002, 04:39:27 PM Don't forget about the PowerPC processors by IBM and Motorola, they have a nice 128 bit pathways, 2 mb level 3 cache, and supercomputer status. Just because it doesn't run x86 apps doesn't mean its not a cool chip.
Posted by FatalDragon [send private reply] at June 11, 2002, 04:55:15 PM But what about the Intel Xeon? Is Xeon just a nickname for PII or is it a different CPU?
Posted by unknown_lamer [send private reply] at June 11, 2002, 05:12:29 PM Xeon is just the server version of the P3.
Posted by AngelOD [send private reply] at June 11, 2002, 05:18:40 PM Xeon is stronger and faster than the PIII. It also has a 1MB L3 Cache.
Posted by metamorphic [send private reply] at June 12, 2002, 06:16:49 AM What about the Itainium? ;) having a 64 bit processor never hurt anyone, even if you do have to sell your body to medical research and your soul to satan to afford it.
Posted by unknown_lamer [send private reply] at June 12, 2002, 10:45:31 AM The Itanium has a nice large number of registers. The IA-64 is better than x86-64 from a technical standpoint (e.g. all commands are the same length), but the x86-64 is a better real world design. Why? Because it is just a 64-bit x86, letting you reuse stuff like optimizations that compilers have worked on for years, whereas with the IA-64 all the years of work people have put in on compilers goes to waste.
Register as a new user