AMD Processors
Decrease font size
Increase font size
Topic Title: Do Processor Performance Benchmarks?
Topic Summary: Intel Vs AMD
Created On: 11/10/2008 06:51 AM
Status: Read Only
Linear : Threading : Single : Branch
Search Topic Search Topic
Topic Tools Topic Tools
View similar topics View similar topics
View topic in raw text format. Print this topic.
 11/10/2008 06:51 AM
User is offline View Users Profile Print this message

Author Icon
weinter
Deprecated

Posts: 321
Joined: 11/08/2008

Does anyone has a link or did a comparison on performance of recent AMD and Intel Processors using the appropriate optimization on GCC?
I think someone mentioned that Windows Benchmarks are not accurate as they are Intel optimized by default so does anyone know a fair benchmark between Intel and AMD?

-------------------------
Acer Aspire 4530
AMD Turion X2 2.1GHZ RM-72 1MB L2 Cache
4GB DDR2 667MHZ
NVIDIA 9100M G
320GB WD BEKT
http://aspiregemstone.blogspot.com/
 11/10/2008 05:53 PM
User is offline View Users Profile Print this message

Author Icon
Firestrider
Senior Member

Posts: 374
Joined: 03/24/2008

I think it is im part because of Intel's C compiler (icc) that is faster when time is of importance.

Here is an article that compares K8, 65nm Core 2 Duo, and K10:
http://www.phoronix.com/scan.php?page=article&item=961
 11/11/2008 10:01 AM
User is offline View Users Profile Print this message

Author Icon
weinter
Deprecated

Posts: 321
Joined: 11/08/2008

Question how many heads(cores) does AMD need to match 2 Intel Core?
Ans : 3

-------------------------
Acer Aspire 4530
AMD Turion X2 2.1GHZ RM-72 1MB L2 Cache
4GB DDR2 667MHZ
NVIDIA 9100M G
320GB WD BEKT
http://aspiregemstone.blogspot.com/
 11/16/2008 07:20 PM
User is offline View Users Profile Print this message

Author Icon
Firestrider
Senior Member

Posts: 374
Joined: 03/24/2008

If an application is well threaded an X3 will beat a Merom or Penryn Core 2 Duo at the same clock speed.

But for single-threaded applications Core 2 Duo beats K10 +5.9% with Merom, and +24% with Penryn given same clock speeds.

AMD really needs to start working on their instruction set, code paths, and core architecture instead of just adding more cores and cache to compete with Intel.

Most of the speedup from K10 to Merom and then to Penryn is the instruction sets SSSE3 and SSE4.1.

The future looks bad for AMD too since AFAIK SSE5 will be their next instruction set introduced in 2011 and the AVX instruction set (which is direct competition and probably better) from Intel will be released sooner with Sandy Bridge.
 11/17/2008 05:02 AM
User is offline View Users Profile Print this message

Author Icon
monte84
Elite

Posts: 2174
Joined: 10/06/2003

I dont quite think AMD is "adding more cores and more cahce" to compete. To be honest I dont think they are going for best performace, but all around value. Intel is the one known for thorwing serious cache into their CPU's. Deneb will be AMD's first real cahce monster, with6MB of L3, something Intel had a long while ago. Athlon X2's had no L3 cache, core2 did, Phenom had a mere 2mb while the core2quads Q6600 had 2x4mb of L2 cache and a Q9550 coming in at a whoping 12MB of L2 cache. Phenom has 512K cahce per core, or 2MB for a quad core. So yeah........... AMD's design of Phenom was meant more towards servers, as the Opteron, where it is extremely competive offering a great performance and minimal power draw compared to the Xeon with its high power draw, mostly for use of fb-dimm memory.

-------------------------
Phenom II X 1055T @3.64Ghz 1.425V
NB @ 2800Mhz HTT @ 2800Mhz
Gigabyte 890FXA-UD5
8GB Crucial Ballistix @ 1866 8-8-8-24-33-1T
AMD Radeon HD6950 2GB GDDR5 shaders @1536 OC to 840/1300
 01/13/2009 09:50 PM
User is offline View Users Profile Print this message

Author Icon
CoolDude55
Junior Member

Posts: 2
Joined: 12/22/2008

so what is the greate one? betwenn AMD and CELERON?
 07/07/2009 02:04 AM
User is offline View Users Profile Print this message

Author Icon
HippyOnABeachInMexico
Junior Member

Posts: 18
Joined: 07/06/2009

It all depends on hardware and how you use it / what for.
AMD fanboy here.

What version of GCC are you talking about 4.4.* or 4.3.* is eglibc or ulibc anywhere used, using python 2.5.* or 2.6.*, kernel and opernrc version.
If using the basic old gcc4.4.* , openrc5.* python 2.6.2-r*, kernel 2.6.31-rc* you might find AMD showing a very good performance over Intel + a few tweaks AMD has a good chance of winning.

Don't have AMD hardware anymore 955 + motherboard "AMDX4 PII 955 not supported" fiasco so can not test.

Keep in mind that AMD based motherboard + cpu will boot much slower than Intel about 8 secs difference in recent tests I did. This is from boot menu to openrc loading. Faster hd on AMD box than on Intel. Using mcp55 x2 3800 or X4 PII 955 790gx "750/710" chip.

Once up and running depending on optimizations and driver they are the same for general usage.

Can you run the fastest turion x2 64bit at MAKEOPTS="-j16" like you can a laptop with Intel 32bit dual core t2600 2.16ghz for 3 years , highly unlikely using customized gentoo build from scratch or any linux.

Is everyone insane like me using these settings = not most users will never use more than 25% load of the cpu, for me 90% or higher loads or switched off , don't need to waist electricity.

Intel is build for number crunching and has the hyper threading makes the OS think there is double the cores installed and yes this scheduling does add a bit on top for more speed but on avg everyday use the 940 and 955 versus q9550 = same as you do not specifically use one strong part or another of and cpu .

The theory that Intel 8 core new cpu's should be able to do 128 threads is nice and likely will but most users with 8 core Amd will not use it in such a way so the AMD might end up winning.

If you crunch number Intel is leading if you use on general usage AMD rocks your world.

I looked at buy a home pc / number cruncher and someone suggested I look at this.

$12k
The 2x HP DLG165 with 2x six core AMD 2.6ghz opterons per box running at usable 12 cores per box and it's cost for crunching numbers versus 2 boxes of 2x Xeon E5530 with 4 cores / 8 virtual cores per cpu per box or 8 installed cores running at 16 cores usable and both costs $500 about per cpu.
If you -j8 a core or 8x 16 virtual cores and leave some for the system to run you definitely have a winner. Problem is to run Opteron nicely at -j8 or more.

I can build 2 boxes / pc's with "24 cores" AMD total or for the same money build 4 boxes with "64 cores" Intel 8cpu total / 8x quad core / effective 8 cores per cpu for the total same money spends and have a ton of ram. $12k spend and AMD is not on top for what I need.

Amd with this parts = excellent server but okay number cruncher.
Maybe -j4 or -j8 times 24 usable cores this one.
$12k @ 96 actual threads to maybe 192 virtual threads.

Intel = okay server and excellent number cruncher.
Will easily -j8 times 32cpu cores / 64 virtual usable cores this one.
$12k @ 256 threads per cpu or actually 512 threads.

Depends what you need. I can use this easily and still might need more. Or buy half the hardware , safe $6k and have 128 actual / 256 virtual threads.

Imagine once ATI stream works nicely then 2x Ati video cards could knock this $12k system to pieces for $220 or do a good job trying. LOL.
 08/28/2009 05:53 AM
User is offline View Users Profile Print this message

Author Icon
Kevin Davis
Junior Member

Posts: 1
Joined: 08/28/2009

Ah, interesting info about open source. I came across something similar some time back.

Payment Gateway
 09/03/2009 09:33 PM
User is offline View Users Profile Print this message

Author Icon
James Walter
Junior Member

Posts: 1
Joined: 09/03/2009

Originally posted by: HippyOnABeachInMexico

It all depends on hardware and how you use it / what for.

AMD fanboy here.



What version of GCC are you talking about 4.4.* or 4.3.* is eglibc or ulibc anywhere used, using python 2.5.* or 2.6.*, kernel and opernrc version.

If using the basic old gcc4.4.* , openrc5.* python 2.6.2-r*, kernel 2.6.31-rc* you might find AMD showing a very good performance over Intel + a few tweaks AMD has a good chance of winning.



Don't have AMD hardware anymore 955 + motherboard "AMDX4 PII 955 not supported" fiasco so can not test.



Keep in mind that AMD based motherboard + cpu will boot much slower than Intel about 8 secs difference in recent tests I did. This is from boot menu to openrc loading. Faster hd on AMD box than on Intel. Using mcp55 x2 3800 or X4 PII 955 790gx "750/710" chip.



Once up and running depending on optimizations and driver they are the same for general usage.



Can you run the fastest turion x2 64bit at MAKEOPTS="-j16" like you can a laptop with Intel 32bit dual core t2600 2.16ghz for 3 years , highly unlikely using customized gentoo build from scratch or any linux.



Is everyone insane like me using these settings = not most users will never use more than 25% load of the cpu, for me 90% or higher loads or switched off , don't need to waist electricity.



Intel is build for number crunching and has the hyper threading makes the OS think there is double the cores installed and yes this scheduling does add a bit on top for more speed but on avg everyday use the 940 and 955 versus q9550 = same as you do not specifically use one strong part or another of and cpu .



The theory that Intel 8 core new cpu's should be able to do 128 threads is nice and likely will but most users with 8 core Amd will not use it in such a way so the AMD might end up winning.



If you crunch number Intel is leading if you use on general usage AMD rocks your world.



I looked at buy a home pc / number cruncher and someone suggested I look at this.



$12k

The 2x HP DLG165 with 2x six core AMD 2.6ghz opterons per box running at usable 12 cores per box and it's cost for crunching numbers versus 2 boxes of 2x Xeon E5530 with 4 cores / 8 virtual cores per cpu per box or 8 installed cores running at 16 cores usable and both costs $500 about per cpu.

If you -j8 a core or 8x 16 virtual cores and leave some for the system to run you definitely have a winner. Problem is to run Opteron nicely at -j8 or more.



I can build 2 boxes / pc's with "24 cores" AMD total or for the same money build 4 boxes with "64 cores" Intel 8cpu total / 8x quad core / effective 8 cores per cpu for the total same money spends and have a ton of ram. $12k spend and AMD is not on top for what I need.



Amd with this parts = excellent server but okay number cruncher.

Maybe -j4 or -j8 times 24 usable cores this one.

$12k @ 96 actual threads to maybe 192 virtual threads.



Intel = okay server and excellent number cruncher.

Will easily -j8 times 32cpu cores / 64 virtual usable cores this one.

$12k @ 256 threads per cpu or actually 512 threads.



Depends what you need. I can use this easily and still might need more. Or buy half the hardware , safe $6k and have 128 actual / 256 virtual threads.



Imagine once ATI stream works nicely then 2x Ati video cards could knock this $12k system to pieces for $220 or do a good job trying. LOL.


Thanks for sharing this information here as I found this as great source of information for me and it increase my knowledge about open source and I was not aware that AMD based motherboard + CPU boot much slower then Intel based motherboard + CPU as you gave the difference of 8 sec Oh Thanks a lot once again as this information helped me a lot in writing my notes about AMD for my assignment at university. Thanks once again!

Edited: 09/03/2009 at 09:42 PM by James Walter
 12/09/2009 01:25 PM
User is offline View Users Profile Print this message

Author Icon
rocky.fus
Junior Member

Posts: 5
Joined: 12/09/2009

I think it is im part because of Intel's C compiler

-------------------------
carpet cleaning burlington vt
Statistics
112018 users are registered to the AMD Processors forum.
There are currently 0 users logged in.

FuseTalk Hosting Executive Plan v3.2 - © 1999-2014 FuseTalk Inc. All rights reserved.



Contact AMD Terms and Conditions ©2007 Advanced Micro Devices, Inc. Privacy Trademark information