Looking for benchmarking help with a Phenom or PhenomII X6

[kladner]

Originally Posted by mrolle

Any of you people, or cats, who are interested, you could post the following numbers on this thread for your AMD CPU:

1. # of cores.
2. Phenom or Phenom II?
3. Northbridge clock multiplier (for a Phenom II, where you can change it).
4. CPU clock speed.
5. L3 cache bandwidth, and what tool did you use to measure it.
..................

1. 6 cores
2. PhenomII, 1090T (Black Edition), Thuban
3. 17.5 x ~200MHz
4. ~3512 MHz
5. Measured with memtest86+ v2.40:
   L1 57,571 MB/s
  L2 18,581 MB/s
  L3  8,586  MB/s
RAM 4,639  MB/s (16GB, DDR1600 OC, 9-9-9-24)
NB clock 2006 MHz (per CPU-Z)

[mrolle]

Originally Posted by pepi37

I have two AMD Phenom x4 960T. One cannot be unlocked, but second is unlocked to X6 . Booth are clocked to 3.7 GHZ.
On x4 and x6 it was enough to just raise CPU core voltage and I am sure they can go above 4 GHZ without any problems: but for now: I am quite happy wit 3.7 GHz. I would like to help but since you attached source program, but not program itself, can you compile it?

Originally Posted by kladner

5. Measured with memtest86+ v2.40:

I thought you all would want to have the source code so that you would trust that it wasn't doing anything bad.

Here's the executable, which runs on Windows XP 64 Bit and Windows 7 64 Bit. (SORRY, it's in the next post)

Also you can use memtest86+ as kladner did. I'm going to check this out myself.

[mrolle]

Originally Posted by mrolle

Here's the executable, which runs on Windows XP 64 Bit and Windows 7 64 Bit.

Sorry I forget the attachment. Here is is.

[kladner]

Originally Posted by mrolle

I thought you all would want to have the source code so that you would trust that it wasn't doing anything bad.

Here's the executable, which runs on Windows XP 64 Bit and Windows 7 64 Bit. (SORRY, it's in the next post)

Also you can use memtest86+ as kladner did. I'm going to check this out myself.

Please don't take this as in any way accusative. I'm just passing it along. When I downloaded the memory test, Norton Internet Security jumped on it and deleted it. As near as I can make out, this is based on their reputation analysis, and that this software has no real reputation. This is hardly a surprise for a low volume, private distribution.

Here is a link to the Symantec explanation.
http://securityresponse.symantec.com...051308-1854-99

There is a "Removal" tab which allows dispute of this classification. However, I am reasonably confident that this is a false positive, so I will work around Norton. I am curious if Michael's app gets different values.

EDIT: I tried to block Norton's response to this file. I got as far as being able to save the download. When I tried to run it, Norton jumped in again and deleted it.

Here's the report I got out of that:

Full Path: c:\users\kieren\desktop\mem test.exe
Threat: WS.Reputation.1
____________________________
____________________________
On computers as of 2/27/2012 at 7:26:00 PM
Last Used 2/27/2012 at 7:47:16 PM
Startup Item No
Launched No
____________________________
____________________________
Very Few Users
Fewer than 5 users in the Norton Community have used this file.
____________________________
Very New
This file was released less than 1 week ago.
____________________________
Medium
This file risk is medium.
____________________________
Threat Details
Threat type: Insight Network Threat. There are many indications that this file is untrustworthy and therefore not safe
____________________________
http://www.mersenneforum.org/attachment.php?attachmentid=7716&d=1330391109 Downloaded File Mem Test.exe
Threat name: 
WS.Reputation.1 from
mersenneforum.org

mem test.exe
____________________________
File Thumbprint - SHA:
fa67d5ec839031f139e5e98afa3b6103310a779c011f2839248234403eb21de5
____________________________
File Thumbprint - MD5:
805a6c78433fda4a6565fe49777bffb3
____________________________

Again, I am just passing on information. I don't really believe Norton on this one, but it is being very persistent.

[Prime95]

Originally Posted by kladner

I don't really believe Norton on this one, but it is being very persistent.

Dang, that's one annoying piece of software. It reminds my of Windows Vista -- constantly in my face with "Are you sure you want to do this?" messages.

[LaurV]

I love "Are you sure you want to do this?" messages

To be on topic, I dlded the exe on some AMD computer on my job, and eset's not32 (which I consider the best antivirus program novaday, and not only me if you check the virusbuletin site) did not jump on it. Unfortunately as I found later, I can't run it because the os is xp32... well I was a bit stupid here... I don't have other computer with new AMD's around. Sorry.

[mrolle]

I got memtest86+ from memtest.org. What you get is a bootable ISO file, which you can burn to a CD and then boot from it.

It is intended for stress testing your memory and looking for errors. However, it also shows the speed of L1, L2, L3, and memory in MB/s. For my purposes, this is nice because what I actually wanted to know was the L3 speed, independent of the CPU speed.

From the data from kladner, above, and running my own tests, it appears that the L3 speed doesn't vary with the CPU speed or the number of cores, and it is about proportional to the NB clock speed. It works out to somewhere around 14 NB clock cycles to get one line from L3.

[mrolle]

I'm satisfied now about the L3 cache speed, so I'm not looking for any other benchmarking help at the moment. Thanks to all you who are willing to help, and I may call on you in the future.

On another topic, I'd like to see a discussion about speed of AMD vs Intel, and how that would influence your future computer purchases. It seems that AMD is generally considered (at least in the GIMPS community) to be somewhat slower. But what difference would it make to you if AMD became faster?

Might there be a more suitable forum for such a discussion? Any suggestions? Thanks.

Michael

[kladner]

Originally Posted by LaurV

I love "Are you sure you want to do this?" messages

[SNARK]Yeah. They're a regular barrel of laughs. [/SNARK] Win7's distrust of most legacy software is pretty annoying, too.

Progress report: I shut off Norton Anti-virus and tried double-click run on Mem Test.exe. I got a DOS box with the cursor blinking one line below the prompt. I tried again running Mem Test from a command prompt (after renaming it "mem_test"), and got the same thing. On the bright side, Malwarebytes Antimalware sees nothing wrong with the executable.

Is there anything else I should be doing to run this in Win7-64? Should I have waited longer than a couple of minutes for it to complete?

PS01: I just tried running it again, forgetting that the Norton disable had expired, and the file got deleted again.

PS02: @mrolle- I just saw your most recent post, so I'll stop trying to run Mem Test. (Unless you have suggestions for making it run. I'd still be interested in comparing its results with those of memtest86+.)
memtest86 has been around for a long time, and I believe that it is highly regarded in enthusiast/overclocker circles. In many ways it compliments Prime95 Torture Tests in that it helps pinpoint memory problems. I see these two programs as real pillars of establishing that a system is stable and reliable.

[Dubslow]

Originally Posted by mrolle

I'm satisfied now about the L3 cache speed, so I'm not looking for any other benchmarking help at the moment. Thanks to all you who are willing to help, and I may call on you in the future.

On another topic, I'd like to see a discussion about speed of AMD vs Intel, and how that would influence your future computer purchases. It seems that AMD is generally considered (at least in the GIMPS community) to be somewhat slower. But what difference would it make to you if AMD became faster?

Might there be a more suitable forum for such a discussion? Any suggestions? Thanks.

Michael

It's fairly consistent across the internet, fan boys ignored. There's lots of benchmarks out there, try e.g. anandtech.com or tomshardware.com, they have lots of reviews and benchmarks. The latter at least usually tests around 10 CPUs at a team, split between AMD and Intel, whenever a new release is out. (Tom's Hardware Bulldozer Review, for example). If you go through the Sandy Bridge review and the one I just linked, that's where I usually get my reference numbers. Of course, Prime95 throughput is more or less consistent with what the review benchmarks would lead us to expect. If AMD gets their crap together and pulls ahead, then for me at least I'll buy AMD. (Note that also in the early 2000s the situation was reversed, where AMD had a sizeable performance lead over Intel, so another flip flop seems entirely within reason.)

[kladner]

Originally Posted by Dubslow

[snip] If AMD gets their crap together and pulls ahead, then for me at least I'll buy AMD. (Note that also in the early 2000s the situation was reversed, where AMD had a sizeable performance lead over Intel, so another flip flop seems entirely within reason.)

It is possible that AMD could pull off another upset of Intel. But the money, literally, and the die size edge, are on Intel's side.

I have used AMD processors for more than ten years. But except for the first of about 3 computer builds in that period, my choice was driven by economic factors rather than absolute performance. Now that I'm more heavily involved in GIMPS, I realize that relative power consumption should have played a bigger role in my selection. An i7 x4 would have drawn noticeably less power that the 1090T, overclocked or otherwise.

On the other hand, there's no way that I could afford a x6 i7. As things stand, I can give two cores of the Phenom II to drive mfaktc, and still have three for P-1 and one for LL/DC (I'm alternating on the last pair.)

Absolute performance aside, this box does deliver substantial output. (The 460 GTX plays no small part in that.)