PDA

View Full Version : Tom's Hardware P4 retraction



Himself
22nd November 2000, 15:46
http://www.tomshardware.com/cpu/00q4/001122/index.html

Yawn. http://forums.murc.ws/ubb/smile.gif

Himself
22nd November 2000, 15:56
http://www.hexus.net/14ghz.shtml

Apples to Apples at 1.4GHz.

Pace
26th November 2000, 15:39
He's changed his mind again. A recompilation helps the P4 a lot, as well as the P3/K7 - I've not read it (yet?) though http://forums.murc.ws/ubb/wink.gif

Paul.

Rocky@intelzone.com
26th November 2000, 18:51
The P4 is not made to run on old hardware or with old software. The P4 is meant for the future which means that it score on tests that use old software will not be as good. This is what Intel is claiming at the moment, we will wait the next couple of months to see if this is true.

------------------
Rocky@intelzone.com
Your source for all of Intel's news and views.

Wombat
26th November 2000, 22:39
>The P4 is not made to run on old hardware or with old software. The P4 is meant for the future

On the contrary, the P4 is supposed to be the next big step for intel's x86 line. The "future" is IA-64.

Rags
26th November 2000, 22:43
Take a look at what Himself has bannered as proof.

Tom has updated his benches to show what a simple recompile can do for a P4.

Other sites have seen the same thing in other tests.

Rags

Himself
27th November 2000, 06:39
<A HREF="http://www.amdzone.com/flask.cfm">Amd Zone</A> has an Athlon 550 doing 8fps in Flask.

The bottom line is that the P4's saving grace is supposed to be games, unless Intel is going to bribe developers to provide patches to existing games, it's a moot point.
Not only do you need an entire new system to use the P4, you need an entirely new library of software to take advantage of it. When and if SSE2 finally gets going, then AMD will also have SSE2 support. In the meantime you can buy a $$$$ system to do you pirating a few frames per second faster than the other guy. http://forums.murc.ws/ubb/smile.gif Thanks to Intel's hard work. http://forums.murc.ws/ubb/smile.gif

Rags
27th November 2000, 12:28
It would rock seriously if AMD got some of the extra enhancements on their next core. This is something I would like to see.

Rags

Another thing I would like to see is an Athlon optimized flask.

Taz
27th November 2000, 15:54
Rags, already done see this page...

http://www.amdzone.com/flask.cfm

Kosh Naranek
29th November 2000, 12:49
Tom already has a blurb about this version ... it's close to the P4 in x87 code but not in SIMD where it gets smacked.
It seems like AMDZone and Tom's are using different methods of testing.
But work continue.......

Himself
29th November 2000, 14:42
Rags,

I'd like to see wider paths between L1 and L2 cache, 256 instead of 64 bit. A better branch prediction unit couldn't hurt as well. Other than that, to compete with the P4, it needs more on the motherboard side, the cpu is already talking to the northbridge at 100MHz DDR, the cpu doesn't need to be changed to do the DDR thing, all the cpus would need is lower multipliers to use higher FSB speeds than 200MHz and of course QA that the cpu can talk at higher frequencies, but that's likely an overclocking like issue. As for SSE2, if the cpu companies are going to cooperate on using the same standard and are not going to be all Rambus about it, great, about time.

My $ .02

Rags
29th November 2000, 18:30
Himself,

I agree that it would rock to see the 256 bit L2 and L1, but the problem is they already shitcanned the processor that was to have it, and really it didn't need to have it (2-4MB L2 is rather large). The Athlon in Tbird form probably would benefit little more than a bump with an increased L2 and L1 bitpath, the L1 on the Tbird is absolutely huge, and the L2 is rather large as well. I still think that if programs were compiled for the T-Bird, P3, P4, we would see some rather compelling performance. I still think that we could squeeze more juice from the cores we have already, and the K7 core has a LOT to optimize for.

Rags

Himself
29th November 2000, 20:26
That's just the nature of the PC, we are all using Windows if that's any indication of how little most people care about performance. http://forums.murc.ws/ubb/smile.gif MFC, enough said. http://forums.murc.ws/ubb/smile.gif

Rags
29th November 2000, 21:16
I agree.

DentyCracker
30th November 2000, 09:12
Yeah, if we really cared about performance we'd be using .....BeOS, or QNX hehehehehe

Rags
30th November 2000, 11:44
One thing I have to add is there is a difference between the speed of an OS's responsiveness, and the speed at which it allows applications to run. The cool thing is that it's clear there are things that can still be done to improve performance not only for now, but the future. And this power lies with programmers as well as hardware designers.

Rags

Himself
30th November 2000, 13:02
Responsiveness is generally a result of real multitasking with thin time slices, as well as a properly designed GUI API. Windows doesn't have the former except under winnt, and the latter is not designed for performance but for features. Application speed is related to how well the OS was written and designed WRT file systems and memory handling. Windows was never designed, it just mutated from DOS. http://forums.murc.ws/ubb/smile.gif

AmigaOS was more responsive and smooth at 7MHz than my current PC at 850MHz, it's nutty. http://forums.murc.ws/ubb/smile.gif

Indiana
30th November 2000, 14:53
And if you really cared about the CPU-cycles the OS is wasting then you'd use .... AmigaOS http://forums.murc.ws/ubb/wink.gif

Sorry, couldn't resist - but if you're having two computers standing beneath each other, one running AmigaOS on a 68060@50 and the other Win98 on an Athlon@600, you DO have to start thinking...

Indiana
1st December 2000, 12:49
Oh, Himself beat me on this one... Stupid browser-cache.