mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2003-02-13, 06:07   #1
SalemTheCat100
 
Oct 2002

2·13 Posts
Default AMD Barton core

If anyone running Prime95 on a the new barton core?

What, if any, speed improvement does the 512k L2 cache give?
SalemTheCat100 is offline   Reply With Quote
Old 2003-03-18, 16:10   #2
8 ball
 
Mar 2003

138 Posts
Default

The 512k L2 dosen't make much of an impact with the Athlon as it did with the Northwood PIV.
8 ball is offline   Reply With Quote
Old 2003-03-19, 00:28   #3
ebx
 
ebx's Avatar
 
Aug 2002

101 Posts
Default

Quote:
Originally Posted by 8 ball
The 512k L2 dosen't make much of an impact with the Athlon as it did with the Northwood PIV.
Northwood brought in more than only 512K L2.

On the other hand, will the extra help for bigger fft tables?
ebx is offline   Reply With Quote
Old 2003-03-19, 13:26   #4
xtreme2k
 
xtreme2k's Avatar
 
Aug 2002

101011102 Posts
Default

The reason why 512k L2 on Barton was not as significant than the P4 was that :-

- AMD's L2 has always been slower. Its only 64bit vs 256bit on intels'. Not only it has less bandwidth it has a higher latency.
- AMD's L1 is much larger than Intel. Meaning additional L2 do not have as much impact.
- As far as I know Intel P4's FPU can only use the L2 and not the L1. Adding more L2 is everything to P4's FPU performance.
xtreme2k is offline   Reply With Quote
Old 2003-03-19, 20:12   #5
8 ball
 
Mar 2003

11 Posts
Default

Quote:
Originally Posted by ebx
Quote:
Originally Posted by 8 ball
The 512k L2 dosen't make much of an impact with the Athlon as it did with the Northwood PIV.
Northwood brought in more than only 512K L2.

On the other hand, will the extra help for bigger fft tables?
The only real difference is the .13 core and the extra cache, which could be put on because the .13 die was smaller and thus had room for the extra cache. The extra 256k L2 was the major factor in the Pentium 4's road to some performance gain.
8 ball is offline   Reply With Quote
Old 2003-03-19, 20:28   #6
ebx
 
ebx's Avatar
 
Aug 2002

101 Posts
Default

Quote:
Originally Posted by xtreme2k
- AMD's L2 has always been slower. Its only 64bit vs 256bit on intels'. Not only it has less bandwidth it has a higher latency.
I didnt read the manuals so I cant argue on the numbers. But what are 64bit and 256bit refer to? Datapath to memory? They should not make any difference if your data is already in cache. Only filling the cache with data from memory is slower at narrower bus, same clock. That is bandwidth.
ebx is offline   Reply With Quote
Old 2003-03-19, 20:30   #7
xtreme2k
 
xtreme2k's Avatar
 
Aug 2002

2·3·29 Posts
Default

64/256 is the datapath between the actual CPU core and the L2 cache.
xtreme2k is offline   Reply With Quote
Old 2003-03-19, 20:34   #8
Xyzzy
 
Xyzzy's Avatar
 
Aug 2002

100001000110112 Posts
Default

The AMD has a 64-bit path to L2 whereas the Intel has a 256-bit path...

Some fun articles to read...

http://arstechnica.com/paedia/c/caching/caching-1.html
http://arstechnica.com/paedia/b/bandwidth-latency/bandwidth-latency-1.html
Xyzzy is offline   Reply With Quote
Old 2003-03-19, 21:02   #9
ebx
 
ebx's Avatar
 
Aug 2002

6516 Posts
Default

Not to forget Northwood came 'coincidently' with DDR supported chipsets.

Intel L1/L2 operates on different policy from that of AMD by the way.

I guess what I want to see is exact how much improvement the 512K L2 ALONE brings us. No doubt Northwood represents a huge advancing over Williamete(probably 10% performance wise). It was when Intel regained the performance lead over AMD.
ebx is offline   Reply With Quote
Old 2003-03-19, 21:11   #10
xtreme2k
 
xtreme2k's Avatar
 
Aug 2002

2·3·29 Posts
Default

DDR support was not a 'coincident' :D

It was a step to move away from RAMBUS. I guess Intel has finally see how unworthy rambus is and is finally ditching them. However from a TECHNICAL point of view RAMBUS is not a bad choice. RAMBUS in itself is far superior than the fastest DDR/SDR memory at the time. Just that from a PRICING (and some political) point of view it is a very very poor.

I would love to see what the extra L2 on the Barton brings us. But I must have missed some good reviews on them. Most of them show them operating on DIFFERENT CLOCk and DIFFERENT FSB. Which is impossible for comparison.
xtreme2k is offline   Reply With Quote
Old 2003-03-19, 21:40   #11
ebx
 
ebx's Avatar
 
Aug 2002

11001012 Posts
Default

Quote:
Originally Posted by xtreme2k
64/256 is the datapath between the actual CPU core and the L2 cache.
If a 64 bit bus is enough to feed a CPU at say 2.5GHz, 256bit wont help a whole lot. Remember they both run at cpu clock. I actually noted down the numbers of L1/L2 bandwidth for an xp1900 and a p4 2.2 the other day when I ran memtest. I seem to recall intel is over twice as fast in L2. I will dig the numbers out tonight.

Going DDR was the trend. We dont like the price of RDRAM. We also like to have choices. However, Intel may have planned its ddr support timeing wise so it appeared as a coincidence. Intel was at a point very firm on RDRAM. It sued VIA who offered a DDR chipset for P4. It was just the users, or money of users rather, that pushed them over. ' I cant/dont want to afford rdram and sdram is crawling, so I go AMD who runs on ddr.' Ouch, that hurts!
ebx is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Core i5 2500K vs Core i7 2600K (Linear algebra phase) em99010pepe Hardware 0 2011-11-11 15:18
Dual Core to Quad Core Upgrade Rodrigo Hardware 6 2010-11-29 18:48
exclude single core from quad core cpu for gimps jippie Information & Answers 7 2009-12-14 22:04
Optimising work for Intel Core 2 Duo or Quad Core S485122 Software 0 2007-05-13 09:15
Benchmark: AMD T-bred B and Barton optim Hardware 8 2003-11-24 06:22

All times are UTC. The time now is 20:23.


Fri Jul 1 20:23:36 UTC 2022 up 78 days, 18:24, 0 users, load averages: 1.83, 1.78, 1.84

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.

≠ ± ∓ ÷ × · − √ ‰ ⊗ ⊕ ⊖ ⊘ ⊙ ≤ ≥ ≦ ≧ ≨ ≩ ≺ ≻ ≼ ≽ ⊏ ⊐ ⊑ ⊒ ² ³ °
∠ ∟ ° ≅ ~ ‖ ⟂ ⫛
≡ ≜ ≈ ∝ ∞ ≪ ≫ ⌊⌋ ⌈⌉ ∘ ∏ ∐ ∑ ∧ ∨ ∩ ∪ ⨀ ⊕ ⊗ 𝖕 𝖖 𝖗 ⊲ ⊳
∅ ∖ ∁ ↦ ↣ ∩ ∪ ⊆ ⊂ ⊄ ⊊ ⊇ ⊃ ⊅ ⊋ ⊖ ∈ ∉ ∋ ∌ ℕ ℤ ℚ ℝ ℂ ℵ ℶ ℷ ℸ 𝓟
¬ ∨ ∧ ⊕ → ← ⇒ ⇐ ⇔ ∀ ∃ ∄ ∴ ∵ ⊤ ⊥ ⊢ ⊨ ⫤ ⊣ … ⋯ ⋮ ⋰ ⋱
∫ ∬ ∭ ∮ ∯ ∰ ∇ ∆ δ ∂ ℱ ℒ ℓ
𝛢𝛼 𝛣𝛽 𝛤𝛾 𝛥𝛿 𝛦𝜀𝜖 𝛧𝜁 𝛨𝜂 𝛩𝜃𝜗 𝛪𝜄 𝛫𝜅 𝛬𝜆 𝛭𝜇 𝛮𝜈 𝛯𝜉 𝛰𝜊 𝛱𝜋 𝛲𝜌 𝛴𝜎𝜍 𝛵𝜏 𝛶𝜐 𝛷𝜙𝜑 𝛸𝜒 𝛹𝜓 𝛺𝜔