AM2+... first phenoms... i doubt they'll make a lot of difference. 4 cores were not all that useful for games back then. Phenom2 would, but that's AM3, even if backwards compatible with AM2+...

I have Phenom 9950 BE, 2.6Ghz variant. They have L3 cache, it should help.

I also have the special Phenom II 940 BE 3.0 Ghz just in case, that was made just for AM2+. I will also try running Vista era games also in Vista. That way we will see comparison of Crysis in dx9 vs dx10. I have AM3 Phenom II as well, but those are just for comparison. They should run in AM3.

I have better GPUs available for Phenoms. Phenoms can run 1066 memory natively. No more need for OC. Phenoms have faster HT.

In Far Cry 2 and Need for Speed: Undercover I do see 100% CPU utilization, so there is hope.

Testbench 7:
- Gigabyte GA-MA770-UD3 v2.0
- Athlon 64 X2 6400+ ADX6400IAA6CZ (Windsor, released in 2007) with DDR2 running at 800
- Asus GeForce GTX 480 (factory clocks, 701 GPU, 924 memory, 1401 shader, released in 2010). NVidia driver 197.41

3d mark 2006 breakdown, 1024x768, Athlon 64 X2 6400+, GeForce GTX 480:

• Return To Proxycon - 45 fps
• Firefly Forest - 45 fps
• Canyon Flight - 98 fps
• Deep Freeze - 56 fps

3d mark 2006 breakdown, 1600x900, Athlon 64 X2 6400+, GeForce GTX 480:

• Return To Proxycon - 44 fps
• Firefly Forest - 45 fps
• Canyon Flight - 92 fps
• Deep Freeze - 56 fps

3d mark 2006 breakdown, 1600x1200, Athlon 64 X2 6400+, GeForce GTX 480:

• Return To Proxycon - 44 fps
• Firefly Forest - 44 fps
• Canyon Flight - 90 fps
• Deep Freeze - 55 fps

The gain is about 3-5 fps over standard Athlon 64 X2 6000+ with DDR 750.

Games tested:
- F.E.A.R. (2005) - in 1600x1200 with max settings we get 144 fps average, 59 fps minimum in built-in benchmark.
- World in Conflict (2007) - in 1600x1200 we then get 33 average fps in built-in benchmark with the best visual quality settings. 4 fps improvement over standard Athlon 64X2 6000+. Benchmark is quite extreme and real game works better. Playable.
- Crysis (2007) - in 1600x1200 average fps in GPU benchmark is 49 with 4x full screen anti aliasing. Everything else was set to max. Playable.

We didn't test Far Cry 2 and Need for Speed: Undercover as we already know from Testbench 6 that Windsor isn't going to cut it.

Conclusion about Athlon 64X2 6400+ (Windsor) with GeForce GTX 480 and DDR2 800:
- noticeable improvement in Crysis in average fps and game. We can state Crysis is in the playable range now.
- it is a good choice for Windows XP era games (2002-2006)
- mediocre coverage of Windows Vista era (2007-2009) due to slow CPU, noticeable in games from 2008
- we are unable to improve performance of Need for Speed: Undercover or Far Cry 2 to acceptable levels (inherited from Testbench 6)

Bonus Testbench 8:
- Gigabyte GA-MA770-UD3 v2.0
- OCed Athlon 64 X2 6400+ ADX6400IAA6CZ to 3.3Ghz (Windsor, released in 2007) with DDR2 running at 943. Base frequency 254Mhz, 13.0 multiplier, CPU speed 3302Mhz with stock voltage, HT link 1270Mhz, DRAM frequency 471.7, DRAM divisor 7, DRAM timings 5 5 5 15, DRAM voltage 2.2V
- Asus GeForce GTX 480 (factory clocks, 701 GPU, 924 memory, 1401 shader, released in 2010). NVidia driver 197.41

Interestingly, attempt to use 13.5 multiplier and lower base frequency lead to performance regression even though memory timings were the same as was CPU clock and memory clock.

3d mark 2006 breakdown, 1024x768, Athlon 64 X2 6400+, GeForce GTX 480:

• Return To Proxycon - 47 fps
• Firefly Forest - 47 fps
• Canyon Flight - 102 fps
• Deep Freeze - 58 fps

3d mark 2006 breakdown, 1600x1200, Athlon 64 X2 6400+, GeForce GTX 480:

• Return To Proxycon - 45 fps
• Firefly Forest - 46 fps
• Canyon Flight - 93 fps
• Deep Freeze - 57 fps

Games tested:
- F.E.A.R. (2005) - in 1600x1200 with max settings we get 150 fps average in built-in benchmark.
- World in Conflict (2007) - in 1600x1200 we then get 35 average fps in built-in benchmark with the best visual quality settings. Benchmark is quite extreme and real game works better. Playable.
- Crysis (2007) - in 1600x1200 average fps in GPU benchmark is 52 with 4x full screen anti aliasing. Everything else was set to max. Playable.

Conclusion about OCed Athlon 64X2 6400+ (Windsor) to 3.3Ghz with GeForce GTX 480 and DDR2 943:
- this represents an attempt to squeeze more performance out of Windsor. It is faster than Testbench 6. It will be the fastest Windsor.
- it is a good choice for Windows XP era games (2002-2006)
- mediocre coverage of Windows Vista era (2007-2009) due to slow CPU, noticeable in games from 2008
- no instability was observed

Testbench 9:
- Gigabyte GA-MA770-UD3 v2.0
- Athlon 64 X2 6000+ ADV6000IAA5DO (Brisbane, released in 2008) with DDR2 running at 778.
- Asus GeForce GTX 480 (factory clocks, 701 GPU, 924 memory, 1401 shader, released in 2010). NVidia driver 197.41

3d mark 2006 breakdown, 1024x768, Athlon 64 X2 6000+, GeForce GTX 480:

• Return To Proxycon - 40 fps
• Firefly Forest - 41 fps
• Canyon Flight - 89 fps
• Deep Freeze - 51 fps

3d mark 2006 breakdown, 1600x900, Athlon 64 X2 6000+, GeForce GTX 480:

• Return To Proxycon - 39 fps
• Firefly Forest - 40 fps
• Canyon Flight - 85 fps
• Deep Freeze - 51 fps

3d mark 2006 breakdown, 1600x1200, Athlon 64 X2 6000+, GeForce GTX 480:

• Return To Proxycon - 38 fps
• Firefly Forest - 40 fps
• Canyon Flight - 82 fps
• Deep Freeze - 50 fps

We see a performace regression of 2-4 fps over Athlon 64 X2 6000+ (Windsor), while stock Brisbane has 100Mhz higher clock speed and memory is running faster as well.

Games tested:
- F.E.A.R. (2005) - in 1600x1200 with max settings we get 132 fps average in built-in benchmark.
- World in Conflict (2007) - in 1600x1200 we then get 30 average fps in built-in benchmark with the best visual quality settings. Benchmark is quite extreme and real game works better. Playable.
- Crysis (2007) - in 1600x1200 average fps in benchmark is 45 without full screen anti aliasing. Everything else was set to max. When 4x anti aliasing is enabled, we get 44 average fps. We get 5-6 fps less than stock Athlon 64X2 6000+ (Windsor).

Conclusion about Athlon 64X2 6000+ (Brisbane)
- inferior to Athlon 64X2 6000+ (Windsor) in performance
- lower TDP is not a good tradeoff given good silent coolers are available
- it is a good choice for Windows XP era games (2002-2006)
- mediocre coverage of Windows Vista era (2007-2009) due to slow CPU, noticeable in games from 2008
- if you have a choice, choose Windsor over Brisbane

Bonus Testbench 10:
- Gigabyte GA-MA770-UD3 v2.0
- OCed Athlon 64 X2 6000+ ADV6000IAA5DO (Brisbane, released in 2008) to 3.2Ghz with DDR2 running at 916. Base frequency 247Mhz, 13.0 multiplier, CPU speed 3210Mhz with stock voltage, HT link 1235Mhz, DRAM frequency 458.7, DRAM divisor 7, DRAM timings 5 5 5 15, DRAM voltage 2.2V
- Asus GeForce GTX 480 (factory clocks, 701 GPU, 924 memory, 1401 shader, released in 2010). NVidia driver 197.41

Multiplier 13 worked great in Testbench 6 and Bonus Testbench 8. This testbench is identical to Testbench 6 in settings used for Windsor. We get clock by clock comparison of Windsor and Brisbane architecture.

3d mark 2006 breakdown, 1024x768, Athlon 64 X2 6400+, GeForce GTX 480:

• Return To Proxycon - 42 fps
• Firefly Forest - 43 fps
• Canyon Flight - 93 fps
• Deep Freeze - 53 fps

3d mark 2006 breakdown, 1600x1200, Athlon 64 X2 6400+, GeForce GTX 480:

• Return To Proxycon - 41 fps
• Firefly Forest - 41 fps
• Canyon Flight - 85 fps
• Deep Freeze - 53 fps

At the same clock speed, memory speed, HT link, Brisbane lags behind Windsor by about 3-7 fps.

Games tested:
- F.E.A.R. (2005) - in 1600x1200 with max settings we get 134 fps average in built-in benchmark.
- World in Conflict (2007) - in 1600x1200 we then get 29 average fps in built-in benchmark with the best visual quality settings. Benchmark is quite extreme and real game works better. Very minor performance drop over stock CPU. Benchmark was repeated and findings confirmed. Playable.
- Crysis (2007) - in 1600x1200 average fps in benchmark is 48 without full screen anti aliasing. Everything else was set to max. When 4x anti aliasing is enabled, we get 46 average fps.

Conclusion about OCed Athlon 64 X2 6000+ ADV6000IAA5DO (Brisbane) to 3.2Ghz with GeForce GTX 480 and DDR2 916:
- this represents an attempt to squeeze more performance out of Brisbane. It will be the fastest Brisbane.
- it doesn't reliably beat stock Athlon 64 X2 6000+ ADX6000IAA6CZ (Windsor) and lags behind Windsor OCed to 3.2Ghz.
- it is a good choice for Windows XP era games (2002-2006)
- mediocre coverage of Windows Vista era (2007-2009) due to slow CPU, noticeable in games from 2008
- if you have a choice, choose Windsor over Brisbane
- no instability was observed

So easily ~10% lower performance. Not an insignificant downgrade, in fact easily within what improvements from generation to generation are nowadays.

That's fascinating. Must be annoying to buy a new computer which is slower than one from last year...

And a good reason to avoid brisbane altogether nowadays...

AMD 65nm process was a big flop. AMD originally probably expected clocks about 3.5Ghz or higher. Phenom I also had initially very low clocks and TLB bug. 2008 was an extremely bad year for AMD with the TLB bug and Brisbane. In 2009 AMD switched to 45nm process and never had the time to refine 65nm. 2008 was probably a turning point when many switched to Intel. I had Intel core 2 duo at that time.

Nowadays people should get Athlon 64 X2 6000+ (Brisbane) only for historical value reasons as a curiosity, or if nothing else is available. Athlon 64 X2 6400+ (Windsor) is a clear winner for AM2, with Athlon 64 X2 6000+ being a stopgap solution until a cheap higher clocked brother is found. 6400+ conveniently runs memory at DDR2 800 speed so there is no need to mess with OC settings. Just run it at stock speed and it will last for a very long time. Mediocre for Windows Vista era, but it does allow to play Crysis. Buy them now as they may be too expensive in 5-10 years.

I was on AMD back then and i completely missed whole mess. I got earlier athlon x2 when vista was released, i do not remember exactly - something in 4000-5000+ range. 4400+ perhaps. This upgrade was long overdue since i was using athlonXP 2200+ still.

This CPUs were not bad back then, they were competitive enough in terms of price/performance and offered some benefits like support for more memory. That system had 8GB, since vista kind of wanted that...

Then a few years later i swapped it for phenom2. Those were fun and still competitive too, with overclocking more easily available (plenty of unlocked options), again - less issues with RAM, whole unlocking cores and L3 thing...

Intel's LGA775 has its own issues with memory controller still in chipset, whole bunch of chipsets and some weird limitations with use of single/dual rank sticks etc. E8600 system i currently have does not support more than 4GB of RAM for example. A couple AM2 boards i have support 8GB with no issues, being older...

For me the point where AMD became not competitive and outright bad, forcing me towards intel, was FX...

4GB is a quite low limit. Not really for Vista. Sometimes boards support 8GB but you cannot use that with the fastest memory. E8600 has 6MB shared L2 cache and supports DDR3 1066 which is quite low as well. Windsor not having a shared L3 cache must be hurting it. It's essentially a competition between 45nm and 90nm and having to share data through main memory.

The motherboard I used Gigabyte GA-MA770-UD3 v2.0 had no trouble running 4x 2GB at DDR2 1000 speed with Windsor. Gigabyte claims it supports 16GB but 4GB modules are not sold.

Testbench 11:
- Gigabyte GA-MA770-UD3 v2.0
- Phenom X4 9950 BE 2.6Ghz HD995ZXAJ4BGH (Agena, stepping B3, released in 2008)
- Asus GeForce GTX 480 (factory clocks, 701 GPU, 924 memory, 1401 shader, released in 2010). NVidia driver 197.41
- 4x 2GB DDR2 800 unganged running at 5 4 4 12

With Phenom, instead of selecting memory speed, we have the option to select DDR multiplier in BIOS.

4x 2GB DDR2 at 1066 was not stable at SPD settings and Windows XP would crash at boot immediately. 4x 2GB DDR2 1000 was stable with Windsor before, but we never tried 1066. When two memory sticks are present, DDR2 1066 is selected automatically. With 4 sticks DDR2 800 is selected. Two sticks were still not completely stable at autodetected timings. I didn't want to spend more time on this so I just used the same settings I used for Windsor, except we use unganged memory controllers. We want to install Vista/Windows 7 as well and having 8GB RAM is useful. It may be possible to reach higher stable memory clocks by raising the base clock and reducing multiplier, same way we did it with Windsor.

3d mark 2006 breakdown, 1024x768, Phenom X4 9950 BE 2.6Ghz, GeForce GTX 480:

• Return To Proxycon - 40 fps
• Firefly Forest - 40 fps
• Canyon Flight - 90 fps
• Deep Freeze - 50 fps

3d mark 2006 breakdown, 1600x900, Phenom X4 9950 BE 2.6Ghz, GeForce GTX 480:

• Return To Proxycon - 38 fps
• Firefly Forest - 39 fps
• Canyon Flight - 86 fps
• Deep Freeze - 49 fps

3d mark 2006 breakdown, 1600x1200, Phenom X4 9950 BE 2.6Ghz, GeForce GTX 480:

• Return To Proxycon - 38 fps
• Firefly Forest - 39 fps
• Canyon Flight - 83 fps
• Deep Freeze - 49 fps

Results are about in line with Athlon 64X2 6000+ (Brisbane).

Games tested:
- F.E.A.R. (2005) - in 1600x1200 with max settings we get 129 fps average in built-in benchmark, which is 15fps less than Athlon 64 X2 6400+ (Windsor). The number is still high enough not to cause concern.
- Call of Duty 4 Modern Warfare (2007) - in 1600x1200 we get 180-330 fps (195 when aiming at tower with scope). 40-80 fps more than Athlon 64 X2 6000+ (Windsor). This game can take advantage of 4 cores and significantly outperforms Windsor.
- World in Conflict (2007) - in 1600x1200 we then get 34 average fps in built-in benchmark with the best visual quality settings. Benchmark is quite extreme and real game works better. 1 fps more than Athlon 64 X2 6400+ (Windsor) which is within error tolerance.
- Crysis (2007) - in 1600x1200 average fps in benchmark is 45 without full screen anti aliasing. Everything else was set to max. When 4x anti aliasing is enabled, we get 46 average fps. 3 fps less than Athlon 64 X2 6400+ (Windsor).
- STALKER (2007) - in 1600x1200, first mission outside we get about 50-120 fps. No improvement over Windsor.
- Far Cry 2 (2008) - with max settings, in 1600x1200 we get about 35-45 fps in the initial jeep ride with 40 fps most of the time. Not enough for enjoyable experience. This game cannot take advantage of more than 2 cores.
- Need for Speed: Undercover (2008) - in 1600x1200 we get 35-45 fps in the city at the main menu. 50% cpu utilization and lowering resolution does not help. Not enough for enjoyable experience. This game cannot take advantage of more than 2 cores.
- Need for Speed: Shift (2009) - 1600x1200 we get 55-65 fps during race from inside car. No improvement over Windsor.

Conclusion about Phenom X4 9950 BE 2.6Ghz with GeForce GTX 480 and DDR2 800:
- low clocks, should be compared to Athlon 64X2 6000+ (Brisbane) performance wise in games. Does not equal Athlon 64 X2 6000+ (Windsor). Widsor has 1MB L2 cache, Phenom has 512KB just like Brisbane but we get L3 cache.
- very few titles take advantage of more than 2 cores, one of them is Call of Duty 4 Modern Warfare. We didn't solve low performance in Crysis, Far Cry 2 or Need for Speed: Undercover.
- we get 4 moderately fast cores that should improve productivity workloads
- it is a good choice for Windows XP era games (2002-2006)
- mediocre coverage of Windows Vista era (2007-2009) due to slow CPU, noticeable in games from 2008
- AMD failed to deliver improved gaming experience in 2008. There exists Athlon X2 7750 BE released in 2008, but it's just 100Mhz faster, therefore difference will be negligible and it isn't worth getting.
- Athlon 64 X2 6400+ (Windsor) is still the overall winner

Next steps:
- test Athlon X2 7850 BE 2.8Ghz (Kuma, stepping B3, released in 2009)
- it is supposed to be Phenom based, with 2 software locked cores. I wonder if BIOS will be able to unlock them. It is the fastest clocked Phenom X2 released. They are supposed to have L3 cache that we missed in Windsor and Brisbane.
- it is expected to be very close to Athlon 64 X2 6400+ (Windsor)
- further testing is on-hold until I receive this CPU
- the most powerful authentic AM2+ CPU available is Phenom II X4 940 BE 3Ghz which will be tested as well. It is an AM2+ specific CPU and doesn't work in AM3. It is expected to be the overall winner. I do not consider true AM3 CPUs in AM2+ as authentic, they belong to AM3 boards.

That went about as well as i expected. Not only is it no longer AM2, but AM2+, but also fails to be an improvement. Yes, architectural improvements, L3 cache, etc play their role which is evident by ~similar performance at lower clock rate, but also said lower clock rate negates any benefits. And 4 cores are useless. They really needed to make higher frequency dual core one like intel did...

AlexZ wrote on Yesterday, 20:57:

- the most powerful authentic AM2+ CPU available is Phenom II X4 940 BE 3Ghz which will be tested as well. It is an AM2+ specific CPU and doesn't work in AM3. It is expected to be the overall winner. I do not consider true AM3 CPUs in AM2+ as authentic, they belong to AM3 boards.

This seems a little strange to me. Said CPU is based on deneb core, the same as all Phenom2 CPUs apart from 6 core ones (smaller core count ones technically were called differently, but basically were just defective deneb dies with cores or L3 disabled). What's the difference between any other phenom2? The fact it has been packaged differently?

I'd say anything that can run in AM2+ is "authentic". It is not AM2 though, that's different.

But anyway it should perform the same as any phenom2 at the same frequency, and since it is BE... And yeah, there will be an improvement here.

Since we are talking about AM2+/AM3 now ran 3dmark06 a few times on my system, just for comparison:

No crysis this time because this is on win7 x64 and would not be comparable...

And this is AM2+ with DDR2:

Fun fact: it is still slower than E8600...

Archer57 wrote on Today, 00:29:

That went about as well as i expected. Not only is it no longer AM2, but AM2+, but also fails to be an improvement. Yes, architectural improvements, L3 cache, etc play their role which is evident by ~similar performance at lower clock rate, but also said lower clock rate negates any benefits. And 4 cores are useless. They really needed to make higher frequency dual core one like intel did...

4 cores will be useful later, in Windows 7 era with DirectX 11 but 2.6Ghz clock is probably too low for those games. I will add a few more 2008/2009/2010 era games into the original post. I will not be testing the next CPU for about 3 weeks, but can test other games and effect of Windows Vista vs Windows 7 on games. Faster OS will win. Also I need to be able to do TRIM somehow so that also matters.

Archer57 wrote on Today, 00:29:

This seems a little strange to me. Said CPU is based on deneb core, the same as all Phenom2 CPUs apart from 6 core ones (smaller core count ones technically were called differently, but basically were just defective deneb dies with cores or L3 disabled). What's the difference between any other phenom2? The fact it has been packaged differently?

I'd say anything that can run in AM2+ is "authentic". It is not AM2 though, that's different.

It's mostly about package and optimal platform for those CPUs. Phenom II X4 940 BE has no choice of other sockets. AM3 Phenom II will be better served in an AM3 board where it was originally intended. I do have Gigabyte GA-870A-UD3 v2.1 which is an AM3 board that doesn't support Visheras so it's perfect for Phenom II. We could look into the benefits of using DDR2 800 vs DDR3 1600 on games.

Phenom II X4 940 BE should be able to beat Athlon 64X2 6400+ (Windsor) and land below your Phenom II X4 3.2Ghz.

What we learned is that AMD CPUs became obsolete next year, if they were not already obsolete on release date. Intel had some headroom.