Legit Reviews

Today, AMD is announcing a new flagship processor, the Phenom II X4 965 Black Edition. This is nothing more than a speed bump for the Dragon platform, but all of these speed bumps are starting to add up and the Phenom II X4 965 Black Edition processor runs at an impressive 3.4GHz. The performance should scale predictably and the price of the Phenom II X4 965 BE will be $245, which is what the Phenom II X4 955 was back in May 2009 when it first launched.

With the CPU voltage increased to 1.475V in the BIOS we were able to hit 4.2GHz after raising the multiplier up to x21 from x17. This is an overclock of 800MHz, which is a great overclock for using air cooling. This is about 300MHz higher than we got the Phenom II X4 955 processor when we reviewed it a few months ago.

Article Title: AMD Phenom II 965 Black Edition Processor Review
Article URL: http://legitreviews.com/article/1037/1/
Pricing At Time of Print: $245 MSRP

Nice and balanced review Nate!

Pushing the 4Ghz limit with air cooling for an X4 is impressive. Going the extra 200Mhz is a bonus.

The numbers on this processor are inching towards the Intel crop at the top, but isn't quiet an i7 jumper yet.

Still an impressive AMD CPU.

It's nice to see the Excel spreadsheet crunching too.

Any way that you can supply us, the average reader, with the Excel benchmark and instructions? I've tried to find it but I haven't been able to locate it on the net.

Have you tried overclocking with a 32 bit OS? It seems that with a 64 bit OS, overclocking a Phenom II is limited to around 4Ghz, while with a 32 bit OS, it can clock higher.

If that is true any speculation to why?

geokilla wrote:Have you tried overclocking with a 32 bit OS? It seems that with a 64 bit OS, overclocking a Phenom II is limited to around 4Ghz, while with a 32 bit OS, it can clock higher.

I think that rumor is BS... I'll ask Sami and Vincent (kingpin) about that later today as they are here at Quakecon. I just had lunch with AMD PR, so I wish I got this question an hour and a half ago.

Clarification please: In the article, aren't all the HT (hypertransport) frequencies overstated by a factor of two?

Remember, since HT is a DDR technology (transfers data on rising and falling edge of clock), 2000MHz == 4000MT/s where T/s means "transfers per second".

Another note: My Phenom II x4 965 BE in two different gigabyte GA-MA790FXT-UD5P motherboards is extremely flaky unless northbridge and hypertransport speeds are lowered from 2000MHz to 1600MHz in BIOS. The problem seems to be even worse when I updated the BIOS from version F5 to version F6. I am running winxp64 if that matters. Also, this happens even when I underclock the CPU to 2400MHz (from 3400MHz) and underclock my 1600MHz DDR3 memory from 1333MHz to 800Mhz (and run it at 9-9-9-24 timing versus its rated 7-6-7-18 timing). Again... I get the exact same results with two of these motherboards and the two latest BIOS versions. Any ideas?

bootstrap wrote:Clarification please: In the article, aren't all the HT (hypertransport) frequencies overstated by a factor of two?

Remember, since HT is a DDR technology (transfers data on rising and falling edge of clock), 2000MHz == 4000MT/s where T/s means "transfers per second".

Another note: My Phenom II x4 965 BE in two different gigabyte GA-MA790FXT-UD5P motherboards is extremely flaky unless northbridge and hypertransport speeds are lowered from 2000MHz to 1600MHz in BIOS. The problem seems to be even worse when I updated the BIOS from version F5 to version F6. I am running winxp64 if that matters. Also, this happens even when I underclock the CPU to 2400MHz (from 3400MHz) and underclock my 1600MHz DDR3 memory from 1333MHz to 800Mhz (and run it at 9-9-9-24 timing versus its rated 7-6-7-18 timing). Again... I get the exact same results with two of these motherboards and the two latest BIOS versions. Any ideas?

Yes the speed ratings are given at their full rated speeds like you noted.

As for the motherboard/CPU issues... BIOS F6 should be golden... Not sure if that is sign of a flaky CPU or motherboard to be honest. When you say flaky what is happening? Have you tired increasing the voltages just slightly?

geokilla wrote:Have you tried overclocking with a 32 bit OS? It seems that with a 64 bit OS, overclocking a Phenom II is limited to around 4Ghz, while with a 32 bit OS, it can clock higher.

32-bit operating systems do clock higher but I believe it is due to them not loading the memory controller as much due to only using the 32-bit extensions. However there isn't a frequency wall with the OS, I've just noticed under LN2 testing that the x64 OS is maybe 50-100MHz lower on maximum stable frequency. For benchmarking right now there is no need to go x64 as none of the tests I've used have gained from having more than the ~3GB limited in 32-bit operating systems.

I know this is an old thread for an old review but I am way behind on tech these days ,real children have taken the place of my former electronic ones when it comes to my $$. My question is...in the test system specs you listed the EVGA 8800 GTS as the video card used but in the caption under the benchies for Crysis and World at War you list the GTX295. Does that mean the Vid cards were swapped only for those benches? If so I was wondering why switch the video card?

Merlin wrote:I know this is an old thread for an old review but I am way behind on tech these days ,real children have taken the place of my former electronic ones when it comes to my $$. My question is...in the test system specs you listed the EVGA 8800 GTS as the video card used but in the caption under the benchies for Crysis and World at War you list the GTX295. Does that mean the Vid cards were swapped only for those benches? If so I was wondering why switch the video card?

not a very good way to test a CPU if your GPU is forcing framerates below 60 regardless of platform. and it's world in conflict, not world at war
(for regular cpu benchmarks GPU doesnt matter at all, including 3DM06. also it is likely most of the cpu-specific benchmarks were done long before the Phenom II came out, and thus previous results from the 8800GTS were used, you have absolutely no idea how much time goes into running the tests, and flagship CPUs means launch-minute reviews with a very limited amount of time to do them, so reusing results is a must)

Cool I get it. I was mostly hoping to see the results with the GTS8800 since that is the vid card I have. Its not often that items in my setup match whats being reviewed.