CPU scores below avrg. Broken after hard use?

[B-Tec]

Hello.
Finally signed up here after couple of years playing with numbers in BIOS.

CPU is Phenom II 965 BE (C2, 140W). It fails really bad in some of the tests Passmark contains for a CPU. At any different OC setup. The most important scores are -50% or more off when comparing to a same CPU with same OC level. Integer math, Floating point math and Find prime numbers are always low while other scores are usuallu corresponding, if not better.

Can this be the result of keeping the CPU at its limits for 2-3 years? Used to keep it at 3.95-4.00Ghz with 1.5-1.5125 Volts, 36C Idle, 54C full load. Playing games doesent take it over 50C.

Tried benchmarking it with several different OC setups and also a stock setup, wich didnt give me less disappointing results.

-Phenom II 965, C2, 140W.
-Sabretooth 990fx rev1.
-Corsair XMS3 1600Mhz @ 1000mhz cl5 - 15XXmhz cl7.
-Ati 7850 GPU.

Bought the MOBO after Crosshair III stopped working a time ago. I do realize I could do a massive upgrade updating the CPU.

 

[EarthDog]

Nobody here is familiar with passmark really...

 

[B-Tec]

But do CPUs act in such way if kept very OC'd for a long period of time? Huge performance drop? Didnt bench it actually until now, so I cant compare it to how it was at first. Hard to tell by FPS also, as I had a GPU upgrade and got CPU better OC'd.

 

[EarthDog]

No..it works or it doesn't.

 

[Johan45]

Where do you have your NB frequency? Get it up to around 2400 if U can. And speed up that ram if you can. That should improve your score.
Oh and to the forum

 

[B-Tec]

Where do you have your NB frequency? Get it up to around 2400 if U can. And speed up that ram if you can. That should improve your score.
Oh and to the forum

Had NB at 2750mhz and anything below that also. Memory speed used to be around 1550Mhz with 7-7-7-23 timings. Now set to 1000mhz 5-5-5-15 temporary.

Forgot to mention I only have currently 2x2Gb of memory. Doubt this would bottleneck any benchmarks, but could it? Those low scores represent "CPU throughput" as the Passmark site explains. Or at least for "Integer Math" it said.

Can there be compatibility issues with the MOBO, as its AM3+ and probably made for the FX CPUs?

 

[Johan45]

The memory wouldn't be a bottle neck and I'm using the same board with my 965 in my HTPC/SETI machine.I think this board works very well compared to my old one M3NHT which only had DDR2.
I'v never been a big fan of benchmarks accept to find the best settings to squeeze a little more out of my vid card.
With a NB like that your performance should be quite good. Maybe OC the ram and check that it's in dual channel mode.

 

[B-Tec]

Phenom II 965 memory controller cant handle 1600mhz. I was suprised it could do 1550mhz while being OC'd to 3.95mhz with 2750mhz NB.

Another thing that came to my mind is... The DDR sticks are in slots 2 and 4. Memory sticks are high profile and this is the only way to fit them with the CPU cooler. The two slots they are in are however the same color and paired.

Memory is running dual channel.

 

[Johan45]

That's the right position for the sticks anyways. I know that the 965 isn't rated for anything above 1333 but with your nb way up there I thought it might handle it. I've got 16Gb at 1600 and NB around 2500. 50% difference is that just posts you've read where mine does so and so. And you never benched it before so you really have no baseline comparison. If all is well I wouldn't worry about it too much. For me as long as I can Game and rip Bluray I'm happy.

 

[B-Tec]

Cant be happy as theere are lots of baselines for this CPU to compare and in different OC levels. Downlodable with the program or seen an average score of a stock version, wich is just a bit less than my very OC version. As stock setup I score 3.3k overall, while average score is 4.333. OC and I get it to 4.700. Check details and its only the particular, important tests that went very wrong BUT other scored good. Integer math for example is never more than 50% less than the a baseline. Tried different baselines as well.

 

[Mandrake4565]

B-tec comparing your system to another system is very tough. Some benchmarks take more then just the cpu into account. You memory being at 1000mhz could be one issue. Also having your NB Frequency at 2750, is that while your memory was at 1000mhz? Denub's IMC not being able to hangle ddr3 1600 is a myth. Some may not be able to but I have one that can do 1800mhz on the ram. What I feel you should do is, run some tests from future mark and then compare those tests to others that have the same cpu and gpu you have. Look at their cpu scores and see how they compare to yours. The frequency you are running your memory at will have some impact on the tests so you'll have to take that into account also.

 

[Johan45]

Thanks for stepping in Mandrake

 

[B-Tec]

Ram wasnt 1000mhz at the best scored test. It was 15XXmhz, around 1550 anyway and CL7 and many other setups. Also NB. Was that 2750mhz or so. Will try future mark.

 

[trents]

Phenom II 965 memory controller cant handle 1600mhz. I was suprised it could do 1550mhz while being OC'd to 3.95mhz with 2750mhz NB.

Another thing that came to my mind is... The DDR sticks are in slots 2 and 4. Memory sticks are high profile and this is the only way to fit them with the CPU cooler. The two slots they are in are however the same color and paired.

Memory is running dual channel.

It likely can handle 1600 mhz if you relax the timings a bit. Not all Deneb core CPUs will handle the memory at 1600 mhz but many will if the timings are relaxed a bit.

What would be helpful is if you would install CPU-z and attach pics of these three tabs: CPU, Memory and SPD.

To attach images using the built-in forum tool, first crop and save the images to disc. Snipping Tool in Windows Accessories is great for this. Then click on Go Advanced at the bottom of any new post window. When the advanced post window appears, click on the paperclip too at the top. This will load the file browser/uplink tool and the rest will be obvious. You can attach up to three images per post.

 

[RGone]

I agree with Johan45 in that there is no baseline on "your" system since you did not run Passmark when you built the system originally. The fact other similar systems do better today does not tell us that your system has been slow/low in cpu score most likley from the outset.

How Passmark is running the cpu bench today maybe slanted due to hardware of today versus hardware of a few years ago. The latest Passmark as I understand it uses 'physics' in determining the cpu score. Your older system probably does not like that method.

Seems the number one complaint about low cpu score is most likely caused by Passmark not using all the cores of the cpu and must be changed in the program preferences as I understand it. This from g00gling low Passmark CPU scores. I don't nor am likley to have Passmark.
RGone...

 

[Johan45]

^+1 nice find Rgone

 

[B-Tec]

Will get some other benchmark software later. 3Dmark? Just pointing out that there are loads of baselines, wich all have logic scores. Weird is that highest OC phenom baseline is this 3.8ghz. On the benchmark picture I benched with full OC setup (3.92ghz CPU, 2770Mhz NB and 1540mhz CL7 ram. Also updated a bit my sig to this.

The details of the benchmark scores are the strange thing, not only the low score in total. The preferences was also set correctly for 4 cores. This total score was 4.250, wich is less than a stock average, but I once benched 4.7k and anything between that.

 

[Mandrake4565]

I do not know enough about the test you ran to comment. I'm more familiar with Futuremark benchmarks. There is a section on their website where you can search, comparable submissions.

 

[Johan45]

I get a feeling that the results you are seeing( generic 965) are more than likely based on the C3 version which overall was a better chip. AMD made quite a few improvements between the two such as 125W vs140W also memory handling was improved which would translate to better CPU scores in the floating point and pi areas.
I notice that tyhe rest of the scores are quite good this is what leads me to think the way I do. I have the the C3 version, but have never done any benches on it so I couldn't say that it would perform more like the ones in your graphs but I strongly think it would possibly even better with the OC I have on it.

 

[B-Tec]

If the difference between C2 and C3 CPUs would be so big, I would think they had renamed at least the 965 numbers and made it a new CPU. But I also think you are making a point. Just hard to believe the difference if that is why it is.