 |
| |
|
|
Asher
|
 |
Calgary, Alberta
Nov 1999
time: 22:26
|
|
Well, The A7N8X is the Nvidia-based motherboard, right? They are pretty speedy, but there's also a lot of compatibility problems. It's only Nvidia's 2nd motherboard chipset ever, afterall...
I'd prefer to go with an Intel chipset over Nvidia, ALi, Via, or SiS myself, but I've heard both Via and Nvidia aren't too bad these days. 
As for the Radeon9xxx, they've been for sale for a long time now for the most part. The 9700 is about 6 months old, same with most of the others, the 9800 is the one not yet for sale (but will be soon), it was just announced on the 6th.
The 9700 Pro is one of the most expensive ones. FutureShop has it for $500CDN: http://www.futureshop.ca/catalog/pr...id=&newdeptid=1
The 9500 Pro (128MB) is $369CDN at CompuSmart: http://www.compusmart.com/product.a...erPartID=557035
|
|
| |
|
|
|
OneFootInTheGrave
|
 |
Kuzelj
Nov 2000
time: 05:26
|
|
Like here
http://www.anandtech.com/cpu/showdoc.html?i=1783&p=12
well you can see that the athlons are a few speed grades faster than PIV's ... even with the + ratings... that is UT2003 engine on which many new games are on like Splinter cell etc... anyway it does seem that where computation power is required Athlons are better and where bandwidth is the benefit, PIV's rule, adaptation of SSE2 will see PIV's gain much more speed however this is very unlikely for games so... go AMD
and for the radeons, like Asher said 9700's have been for sale for 6 months or so, and 9500 lhave been for sale for 3 months at least... howevr 9800's and 9600's should come to stores by the end of the month... still 9500 are already DX9 capable and 9500pro is faster than GF4 ti4600/4800 and is a DX9 part so do not look for less than that it will be money well spent.
as for the chip I would risk going for an AMD 1700+ as they are well known to go up to 2000 mhz if you just change the multiplier in the bios without changing the voltage and such (just make sure it is thoroughbread B copre that you are buying), but perhaps needs a bit better air cooler... these chips are around 60$ or so... and you can run them at at least 166 mhz fsb or maybe even 200 mhz.. as you will buy Nforce2 chipset and PC 3200 memory... even if you have it at 1800 mhz and FSB at 400 it should be faster than regular 2000 mhz chip at 266 FSB... (and that is 2400+ chip) anyway, just to tell you that there are ways to go about this computer.... anyway that would leave you some spare money for a Radeon 9700 non pro which will be a kick ass card for a long time to come.
Last edited by OneFootInTheGrave on 10-03-2003 at 05:12
|
|
| |
|
| |
|
| |
|
|
|
Asher
|
|
Calgary, Alberta
Nov 1999
time: 22:26
|
|
quote:
Originally posted by OneFootInTheGrave
well you can see that the athlons are a few speed grades faster than PIV's ... even with the + ratings... that is UT2003 engine on which many new games are on like Splinter cell etc... |
Splinter Cell is based on Unreal Warfare, the UT2003 is also based on Unreal Warfare...if anything, Splinter Cell is optimized for the Pentium III more than anything else. 
UT2003 isn't the only game out there, obviously. Q3 and Q3:TA run quite a bit faster on P4s, as does games like Comanche 4, Jedi Knight 2, etc. And 3DMark, if that matters.
Out of the 5 "gaming" benchmarks in the Anand article you linked to, the Athlon won 1.
quote:
anyway it does seem that where computation power is required Athlons are better and where bandwidth is the benefit, PIV's rule, adaptation of SSE2 will see PIV's gain much more speed however this is very unlikely for games so... go AMD |
I'm very confused.
Why would SSE2 be unlikely for games? Over half of the SSE2 instructions were made explicitly for gaming...3D transforms with 64-bit precision, that kind of thing. Seeing as AMD's Athlon 64 will support SSE2, its adoption will only be more widespread now.
Video card drivers also extensively use SSE2.
Seeing as John Carmack has been saying how great HyperThreading is, you can expect the DOOM III engine will be optimized for HyperThreading as well as SSE2, which is why I think seeing as the P4s are already generally faster for gaming, the gap will only grow larger with new apps...
Not to mention how well the non-gaming apps on P4 are, like encoding.
http://www.anandtech.com/cpu/showdoc.html?i=1783&p=18
Notice how a 3.06GHz P4 renders 97.9fps converting a MPEG2 (DVD) file into DivX, while a 3000+ Athlon does it at 71.1fps. Quite a big difference.
Or in Windows Media Encoder where the 3GHz P4 takes 8 minutes to encode a video, the 3000+ Athlon takes 10.23...
Or in 3DStudio, where a 3GHz P4 takes 169 seconds to render a scene, the 3000+ Athlon takes 227...
Or in Lightwave, 3GHz P4 takes 101.5seconds, 3000+ Athlon takes 134 seconds. 
Last edited by Asher on 10-03-2003 at 06:05
|
|
| |
|
| |
|
| |
|
|
|
Asher
|
|
Calgary, Alberta
Nov 1999
time: 22:26
|
|
Look, it basically works like this. At any given time, a minority of the CPU is in use. There's basically a bunch of different "subprocessors" in each CPU. There's the ALU, which does integer/logic operations, the FPU (x87 floating point operations), and the SIMD unit (SSE/SSE2 operations).
A single "thread" (string of instructions) doesn't use all of those segments at once.
What HyperThreading (otherwise known as Simultaneous MultiThreading) does is allow two threads to run on the same processor, rather than one. If one thread is doing AI work on the ALU, another thread can be doing physics calculations using the SIMD and/or FPU units at the same time.
Single-threaded apps show 0% performance increase with it, multi-threaded apps almost always show a boost.
"Load of crap" it is not. HyperThreading 2 on Prescott will be better, though.
|
|
| |
|
| |
|
| |
|
|
|
Urban Ranger
|
|
Apolyton Duke of Off-Topic
|
|
quote:
Originally posted by Asher
What HyperThreading (otherwise known as Simultaneous MultiThreading) does is allow two threads to run on the same processor, rather than one. If one thread is doing AI work on the ALU, another thread can be doing physics calculations using the SIMD and/or FPU units at the same time. |
The very fundamental problem is this: these CPUs are already supposedly "super-scalar," with this executing multiple micro-ops, retiring out of order, branching look ahead, and all that good stuff. So, fundamentally, all the execution units are supposed to be occupied doing something, because of the "smart" controls.
So, when you find some idle execution units in one of them, it can mean one of two things:
1. The control logic isn't smart enough
2. The execution speed outstrips data transfer speed
Neither of these problems can be fixed by "hyperthreading," that's why I said it is a gimmick.
|
|
| |
|
|
|
Asher
|
|
Calgary, Alberta
Nov 1999
time: 22:26
|
|
quote:
Originally posted by Urban Ranger
The very fundamental problem is this: these CPUs are already supposedly "super-scalar," with this executing multiple micro-ops, retiring out of order, branching look ahead, and all that good stuff. So, fundamentally, all the execution units are supposed to be occupied doing something, because of the "smart" controls.
So, when you find some idle execution units in one of them, it can mean one of two things:
1. The control logic isn't smart enough
2. The execution speed outstrips data transfer speed
Neither of these problems can be fixed by "hyperthreading," that's why I said it is a gimmick. |
I don't see why you'd think a superscalar architecture automatically would mean all of the units are being fed?
The problem is the units are all being fed, provided that thread uses the units. A processor today can only use one thread, remember, if that thread is only touching ALU stuff throughout the pipeline, the SIMD/FPU stuff sits idle.
|
|
| |
|
|
|
Asher
|
|
Calgary, Alberta
Nov 1999
time: 22:26
|
|
quote:
Originally posted by Urban Ranger
When I mentioned "execution units," I referred to the ALU's since they are fundamental. FPU is optional.
"Threads" make no difference at that level.
Remember, "superscalar" architecture is supposed to mean multiple micro-ops executed at the same clock cycle because of the hardware parallelism. This should translates to all execution units (ALU's) in use every cycle. If this is not the case, they should fix this problem. |
I'm still confused why you think that superscalar translates to all execution units in use every cycle. Even if it were the case that all ALUs were in use, it's not the case that the FPU and SIMD units are in use, which HyperThreading helps fix.
HT is one way to increase ILP, that's why it's there. And it does work in the realworld in most cases, will be improved in Prescott (Prescott doubles much of the logic on the chip in key areas, as well as L1/L2 cache), and adds instructions for thread syncs.
|
|
| |
All times are GMT. The time now is 05:26.
Apolyton Time is 00:26. |
top of page
|
| archivepost |
|
Forum Rules:
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is ON
vB code is ON
Smilies are ON
[IMG] code is ON
|
|
|
|
|
|
10-03-2003 02:12
|