View unanswered posts    View active topics

All times are UTC - 6 hours





Post new topic Reply to topic  [ 24 posts ] 
Go to page Previous  1, 2

Print view Previous topic   Next topic  
Author Message
Search for:
 Post subject:
PostPosted: Thu Nov 15, 2007 9:28 am 
Offline
Joined: Wed Jan 25, 2006 1:03 pm
Posts: 240
Location: Shakopee, MN USA
marc.aronson wrote:
TVBox wrote:
From what I have read a 3800+ dual core is nothing more that two 1900 cpu's together on the same piece of silicon. Also the sharing mechanism may make them slower than a single core of the same type on single loads.

If all you where doing was transcoding that single core 3700 probably would be faster.


I am not as familiar with AMD, but my intel 3ghz dual core has two 3 ghz cores, not two 1.5 ghz cores. I would be interested in reading more on the AMD line -- can you point me to the article / material that explains the an AMD 3800+ is actually 2 1900+ cores?

I find multi-core very beneficial for my box, as I frequently have multiple tasks running concurrently. It is especially helpful when running CPU-intensive jobs concurrently with doing hidef playback.

Marc


AMD has some good information about their processors and their archetecture here:

http://www.amd.com/us-en/Processors/Pro ... 41,00.html

From the description, they are two processors on one die and have their own independent caches. The thing they share is the access to the memory and what is called a "Hypertransport Bus".


Top
 Profile  
 
 Post subject:
PostPosted: Thu Nov 15, 2007 12:38 pm 
Offline
Joined: Fri May 21, 2004 11:55 pm
Posts: 1206
Location: Silicon Valley, CA
TVBox wrote:
From what I have read a 3800+ dual core is nothing more that two 1900 cpu's together on the same piece of silicon. Also the sharing mechanism may make them slower than a single core of the same type on single loads.

This is an oversimplification of the difference between single and dual core AMD processors. For example, the XP 4000 (single core) runs at 2.6GHz and has 512K cache. The XP2 4000 has two processors which each run at 2.1GHz and has 1M cache. So the dual-core processors have way more compute bandwidth than 1/2 of a single core of the same "rating" (marketing number).

I do not argue that a dual-core processor of the same "number" can be slower than a single-core on a single process. But that does not even begin to represent the actual load a system sees under Myth. Dual-cores are very useful to a Mythbox.

_________________
Do you code to live, or live to code?
Search LinHES forum through Google


Top
 Profile  
 
 Post subject: X1 verses X2 Benchmarks
PostPosted: Fri Nov 16, 2007 1:49 am 
Offline
Joined: Mon Apr 23, 2007 1:45 pm
Posts: 405
Location: Fargo, ND, USA
Yes the two 1900 cpu's together on the same piece of silicon is an over simplification but not by much. It all comes down to how fast your processes run. There seems to be this notion that a dual core 3800+ is two complete 3800 processors. If this where true the dual core 3800 should draw twice as much power as a single core and perform in half the time but they have the same consumption. If I match the processor voltage and use 90 nm cores a AMD x2 3800 draws 65 watts and a AMD x1 3800 62 watts. The cache is twice as big in the dual core and could account for the 3 watts difference. Personally I want to know the Miles Per Gallon difference between doing exactly the same job or in this case watts consumed per job.

Tom's Hardware has a excellent benchmarks site. You can compare 91 diferent 91 AMD and Intel processors with 35 different benchmarks

http://www23.tomshardware.com/cpu_2007.html

Below I have tried to pick benchmarks that list all four of these processors and in some way apply to the Knoppmyth situation. I picked the Athlon 64 5200ee because it is a 65 watt chip and the best bang for the watt in the faster processors. I included the Athlon 64 3200 x1 because that is what I am currently running. In the war of the single core 38 verses the dual core 38 I put a # for the winner

Benchmark 2 minutes of DVD to Xvid
Athlon 64 5200 x2 --------- 02:51
Athlon 64 3800 x2 --------- 03:33
Athlon 64 3800 x1 --------- 03:29 ---- #
Athlon 64 3200 x1 --------- 04:02
less is better

Benchmark: Clone DVD Transcoding DVD9 to DVD5
Athlon 64 5200 x2 --------- 09:10
Athlon 64 3800 x2 --------- 11:18 ---- #
Athlon 64 3800 x1 --------- 10:12
Athlon 64 3200 x1 --------- 12:06
less is better

Benchmark 2 minutes of DVD to Divx
Athlon 64 5200 x2 --------- 02:29
Athlon 64 3800 x2 --------- 03:07 ---- #
Athlon 64 3800 x1 --------- 04:12
Athlon 64 3200 x1 --------- 04:59
less is better

Benchmark Cinema 4D - Rendering from scene
Athlon 64 5200 x2 --------- 04:44
Athlon 64 3800 x2 --------- 06:08 ---- #
Athlon 64 3800 x1 --------- 10:03
Athlon 64 3200 x1 --------- 12:03
less is better

Benchmark Photoshop CS-3 Filtering a 69mg TIF-Photo
Athlon 64 5200 x2 --------- 02:26
Athlon 64 3800 x2 --------- 03:17
Athlon 64 3800 x1 --------- 02:39 ---- #
Athlon 64 3200 x1 --------- 03:13 -----yes this is right
less is better

In the core wars I have seen no instance where the dual core was twice as fast as the single. It looks to me that they pretty much average out to be the same. Even if I compare the the 3200 x1 against the 5200 x2 there was only one instance that the 5200 completed in less than half the time and most of the time the 5200 was only 150% faster. So if I use the AMD's arbitrary designation numbers, 5200 is 160% larger than 3200. Thats also the conclusion that the Benchmarks came up with.

In the "work for watt game" it seems that only AMD players in the game are the
AMD Athlon 35 watt 3800+EE x2
AMD Athlon 35 watt 3500+EE x1

The Athlon 64 5200 x2 65watt is 15 watts to high to be in the same class and even then the idle losses would throw it out of the running. I am unfamiliar with the Intel products and haven't checked them but would like to see some comparisons using the same criteria if anyone out there is willing.

I am still building my 35 watt 3800+EE x2 that hasn't changed but what has changed is that I now have more respect for my old 3200 machine.

_________________
TVBox
LinHES R8.6.1


Top
 Profile  
 
 Post subject:
PostPosted: Fri Nov 16, 2007 8:53 am 
Offline
Joined: Fri May 21, 2004 11:55 pm
Posts: 1206
Location: Silicon Valley, CA
Yes, this is a good analysis. But it's kind of the opposite of what your previous comment implied.

You said before that the 3800 was like two 1900's in a single package. That implies that the dual-core will be HALF as fast on a single process as the single-core processor. In your analysis you are showing that the processors are more alike than different when doing a single task, i.e. the 3800 single- and dual-core chips have very similar times on the single tasks listed.

In a KM box there are many processes which benefit from having a second core. I started with a single-core processor, then went to a "hyperthreaded" processor, and now finally a dual-core. The dual-core chip handles Myth much better.

I have commercial flagging set to run at the same time as the recording. My shows are flagged within a minute or two of the show end. The viewing of shows does not cause stuttering since the processor has another core to dispatch on critical processes while I watch TV on the other. There are many, many examples of where a dual-core chip handles "real" MythTv loads better than a single-core.

_________________
Do you code to live, or live to code?
Search LinHES forum through Google


Top
 Profile  
 
 Post subject:
PostPosted: Fri Nov 16, 2007 9:17 am 
Offline
Joined: Tue Jan 30, 2007 1:27 am
Posts: 299
Quote:
In the core wars I have seen no instance where the dual core was twice as fast as the single.


Your tests seem to focus on accomplishing a single task. On my box I may be running several commflagging and transcoding jobs simultaneously. I suspect this is where the dual core would really shine.

A more useful KM-relevant comparison would be to compare times running a single transcode vs two vs three simultaneously, and then adding some commflagging into the mix.

I actually have an X2 3800 box running windoze, and a spare drive. And my main box is an Athlon 3200. If someone can help me by writing a benchmarking script to launch some transcode jobs and time them, and help me setup a dual boot, I'll load KM on the spare drive and do the tests and post the results.


Top
 Profile  
 
 Post subject:
PostPosted: Fri Nov 16, 2007 7:44 pm 
Offline
Joined: Tue Jan 18, 2005 2:07 am
Posts: 1532
Location: California
Liv2Cod wrote:
There are many, many examples of where a dual-core chip handles "real" MythTv loads better than a single-core.

I concur with Liv2Cod. The dual core chips provide significant benefits in environments in which there are multiple processes running in parallel and my knoppmyth box is an example where I have many processes running concurrently.

Marc


Top
 Profile  
 
 Post subject:
PostPosted: Fri Nov 16, 2007 7:54 pm 
Offline
Joined: Wed Mar 07, 2007 9:51 am
Posts: 173
Location: Uniontown, PA
mogator88 wrote:
Quote:
In the core wars I have seen no instance where the dual core was twice as fast as the single.


Your tests seem to focus on accomplishing a single task. On my box I may be running several commflagging and transcoding jobs simultaneously. I suspect this is where the dual core would really shine.

A more useful KM-relevant comparison would be to compare times running a single transcode vs two vs three simultaneously, and then adding some commflagging into the mix.


This is true. Trying to gauge multicore performance on a single task will NOT show any benefit on a multicore processor. It will only depend on core speed.

My discussion about hyperthreading says about the same thing, but says that hyperthreading is not a TRUE multi-core implementation.

If you want a single process to complete faster....you need a faster core speed. Multiple cores help if you're doing more than ONE process at the same time. The only bottleneck there would be I/O issues, such as PCI, AGP, IDE, SATA, and USB bandwidth.


Top
 Profile  
 
PostPosted: Tue Dec 04, 2007 12:28 pm 
Offline
Joined: Mon Apr 23, 2007 1:45 pm
Posts: 405
Location: Fargo, ND, USA
Did any one order a ADD3800 and actually get one from Newegg? What I have heard of so far is that they received a ADO3800IAA5CS, AKA a standard 65W Windsor.
TVBox

_________________
TVBox
LinHES R8.6.1


Top
 Profile  
 
PostPosted: Mon Dec 10, 2007 11:13 am 
Offline
Joined: Wed Jan 25, 2006 1:03 pm
Posts: 240
Location: Shakopee, MN USA
TVBox wrote:
Did any one order a ADD3800 and actually get one from Newegg? What I have heard of so far is that they received a ADO3800IAA5CS, AKA a standard 65W Windsor.
TVBox


I will see if I get this at Christmas time... :wink:


Top
 Profile  
 

Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 24 posts ] 
Go to page Previous  1, 2



All times are UTC - 6 hours




Who is online

Users browsing this forum: No registered users and 26 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group

Theme Created By ceyhansuyu