AMD Phenom II and Athlon II Performance Comparison

Add Your Comments

The Athlon II X4 series is an excellent low cost alternative to its big brother, the Phenom II X4. I have run several tests and compiled the results in a comparative format. The Athlon II shows some weakness – about -25% averaged in a heavy multitasking environment as compared to the Phenom II, but is within .25% in a non-multitasking situation. For general purpose and gaming oriented computing it is hard to beat, fairly cool running and not the power hog many think it is.

A WARNING: BIOS, BIOS, BIOS!!! Do not assume you need to update, and if you do be prepared to revert. When I first installed this CPU I did not update my bios and I had full functionality except for the core temp sensors reading -254. I decided to update and lost the ability to clock past 3.4 and also the ability to lower the NB multiplier (that bios rev. for some reason is no longer available for download). Many BIOS resets and even a processor swap or two may be necessary to get where you want (you could also get a low clocking CPU and then it is a futile battle). Remember this!! Some modern BIOS actually reduce functionality. This was only a warning not a suggestion. I bear no responsibility if you have a problem; this was a kind of public service announcement.

Power usage and heat

The Athlon II CPU at stock voltage (1.4v) only burns about 18 watts more than the 905e (1.15v), and at the same voltage burns 2 watts less. As to the heat question, the 620 is 2.7* cooler at the same voltage and 3.8* hotter at stock voltage.

Results

Below are the results of a direct comparison of the Athlon II X4 and the Phenom II X4 with all settings the same. The areas in red and green are load/no load, where red is a heavy load and green is no load, and are used in both charts. The first chart is total average and the second is a heavy multitasking to no load average.

Head 2 Head AthII PhII

Head 2 Head AthII PhII

Head 2 Head AthII PhII

So that is a clock for clock comparison.

Overclocking

The overclock is very dependent on the motherboard used; you must be able to achieve a high HT-ref clock!

As to the overclock, if you already have a system and wish to upgrade, test your Ref clock for it’s maximum, keeping in mind that you need at least 308 to reach 4.0GHz. You may need to lower your multipliers (NB, HT and memory) for this. After 3GHz you will probably need to add voltage to the core.

Validation at 3.9 VALID was easy, as well as error free crunching (WCG) and Folding. Stability was another issue. From time to time I had a reboot.

Here are some shots of what I consider a good OC for this CPU.

3.7Ghz

100% load temps after 1 hour Bionic/WCG and F@H GPU 100% cpu useage and F@H CPU 100% useage. All projects @ 100% running concurrently. CPU-Z and Sandra for settings.

100% load temps after 1 hour Bionic/WCG and F@H GPU 100% cpu useage and F@H CPU 100% useage. All projects @ 100% running concurrently. CPU-Z and Sandra for settings. A/C AF64

This simple integer benchmark focuses on the branch prediction capabilities and the misprediction penalties of the CPU. It finds the solutions for the classic "Queens problem" on a 10 by 10 sized chessboard

CPU PhotoWorxx test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.

CPU ZLib test uses only the basic x86 instructions, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.

CPU AES test uses only the basic x86 instructions, and it's hardware accelerated on VIA PadLock Security Engine capable VIA C3 and VIA C7 processors. The test consumes 48 MB memory, and it is HyperThreading, multi-processor (SMP) and multi-core (CMP) aware.

This benchmark measures the single precision (also known as 32-bit) floating-point performance through the computation of several frames of the popular "Julia" fractal.

This benchmark measures the double precision (also known as 64-bit) floating-point performance through the computation of several frames of the popular "Mandelbrot" fractal.

This benchmark measures the extended precision (also known as 80-bit) floating-point performance through the computation of a single frame of a modified "Julia" fractal.

I used 667 DDR2 on a Biostar AM2+ MB for this testing as I was looking at this as a low cost upgrade option for those with little cash and still riding a dually or an older 65nm quad.

As a side note, these make excellent HTPC processors when stock clocked and undervolted:

These temps recorded using Artic Cooling AF64 HSF

I went as low as 1.12v stable.

-Archer0915

Leave a Reply

Your email address will not be published. Required fields are marked *

Discussion
  1. QuietIce
    The only way to really compare is to run the rig long enough (like a week or two) to get some kind of statistically reliable data. If you get detailed results (like SETI provides) then you can shorten that time down to a couple of days if you happen to get the right types of work units to compare. You also have to run the two rigs at the same speeds, including the same RAM speeds, to get good results.

    Also, I don't know about WCG but Einstein will occasionally download a new app (automatically) for crunching. You'd need to make sure both sets of results were run on the same app ...


    Yeah as I said seems faster.

    3297 ppd Phenom II

    3498 ppd Athlon II

    2837 ppd 7750 BE

    As I said without the same WU there is really no way to tell, and what ice said. Setups are in my sig, I think the 620 is at 3.5 ATM. Gotta clean them filters.

    I
    The only way to really compare is to run the rig long enough (like a week or two) to get some kind of statistically reliable data. If you get detailed results (like SETI provides) then you can shorten that time down to a couple of days if you happen to get the right types of work units to compare. You also have to run the two rigs at the same speeds, including the same RAM speeds, to get good results.

    Also, I don't know about WCG but Einstein will occasionally download a new app (automatically) for crunching. You'd need to make sure both sets of results were run on the same app ...
    tom_ozahoski
    Do you know the folding PPD difference between the two? I'm looking to upgrade to a quad soon and wonder if the L3 makes much of a difference.


    I am speaking of, WCG Boinc only. When I was doing the testing it seemed that it was laying out more points than the 905, unless I had the same WU I could not be sure but I have it clocked 300 faster than the 905 and it is turning in way more than a 10% PPD difference, again the WU has a lot to do with that.

    I will see what I can do in the coming weeks about getting you some data.
    Nice write up. I have the smaller brother of that Biostar motherboard you used. It just doesn't have dual PCIe 16x slots, just one.

    Good board. If you OC, get a heatsink setup for the mosfets near the CPU.
    I suspect some of those calculations are quicker on the Athlon because of a lack of L3. Calculations small enough to reside completely on L2 should be quicker on the Athlon since there's no L2 overhead and there's a much smaller RAM latency ... ;)
    QuietIce
    Nice work, Archer - thanks for sharing! :thup:


    Thanks.

    Yeah, I saw a review pointing out deficiencies and I wanted to find the facts. Though correct in their conclusions, they were still misleading as the review made it sound more like across the board issues. I put the 620 and 905 under an unrealistic load to find the real deficiencies.
    Nice work U did, I cmparing the results of athlon x4 620 3,7Ghz to my phenom 955BE c2 3,8Ghz it isn't much gap between the score, only the photoworxxx that mine processor is leaving athlon x4 620 3,7Ghz way behind.

    so I concluding athlon is really nice in performance
    Dolk
    Yeah I hear you on that. All I could get was 3.4ghz :/


    Do you know your max Ref clock? I would suggest lowering all multi's. and testing for that just to see if it is the proc. or the ref clock.
    I was using a http://www.biostar-usa.com/app/en-us/t-series/introduction.php?S_ID=377 for those tests.

    On a http://www.biostar-usa.com/app/en-us/t-series/introduction.php?S_ID=374 I max out at about 3.7, 3.6 stable.

    and on this http://www.msicomputer.com/index.php?func=proddesc&maincat_no=1&cat2_no=171&prod_no=1871 I cant get the Ht-ref over ~260 which really sucks.

    An OC with this bad boy is really dependent on your MB and the memory divider used.