As expected it continues to do terrible in single threaded tests, even worse than I thought. Reaching only 8-12% improvement over zambezi (high end desktop processors based on bulldozer cores).
Well after reading through a few reviews the verdict is out.
The bad:
-Improvement over zambezi in single threaded tests was only 8-12%
-Improvement over zambezi in multithreaded tests was only 12-16%
-No improvement in power consumption over zambezi
-Almost all of the performance improvement over zambezi comes from a clock rate increase (which they were able to do thanks to switching the design over to the same high density logic gate library that their GPUs use) IPC has only increased by 2-6% over zambezi.
-Power scaling with overclocking has only slightly improved over zambezi. It still exhibits huge increases in power consumption from slight overclocks and requires voltage bumps even for minor overclocks. 10-20% OC is the highest you can possibly go on most high end air/liquid cooling setups and that results in another 15% increase in performance. That slight overclock will increase the chips power consumption by more than 50%.
-Intels processors continue to beat the living tard out of it in single threaded tests
-Intels processors continue to beat it by a reasonably significant amount in mixed tests
-Intels processors continue to beat it by a small amount in multithreaded floating point heavy tests
-Intel processors continue to beat the living tard out of it in games (I actually don't fully understand why, what kind of code profile do video games usually fall into?)
-Intels processors continue to use significantly less power both at idle and at full load
-Intels processors continue to offer better performance per watt on average
-Intels high end processors come with an IGP while AMDs do not
-Intels processors continue to OC by higher margins with less voltage increase and with much less power consumption/heat increase
The good
-The clock rate increases did not increase power consumption. Therefore stock power consumption levels are still tolerable.
-It is now without question better than phenom II in all applications not just most.
-They are priced surprisingly low.
-The FX-8350 goes head to head with the 3570K and 3770K in multithreaded integer heavy tests, which most slow applications these days are. It beats the 3570K far more often than it loses in this area. Similar situations can be found in the other models. They consistently beat higher priced sandy/ivy bridge processors in multithreaded integer applications.
-Therefore performance per dollar can be better or much better than sandy/ivy bridge at stock depending on what type of applications you're going to be running.
-We all know that the architectures performance is subpar with certain types of applications. But the gap between AMD and Intel is narrower now for those applications than it was before.
Conclusion: It's still not great, but it competes with Intels low end and medium end offerings fairly well and unlikely zambezi you could actually justify this purchase in some scenarios. It can reach the performance of the 3570K and 3770K in many applications for less money but requires more power to do it and doesn't have as much overclocking headroom available. It's exactly what we all expected as far as the level of improvement goes. What is impressive is that the performance improvement going from zambezi to vishera is greater than the performance improvement going from sandy bridge to ivy bridge was. And Intel had the advantage of moving to 22nm while AMD achieved that greater performance improvement on the same manufacturing process. What doesn't make sense is why AMD didn't use their high density library to begin with, considering most of the performance improvement came from that and this library has been around for years I don't see why it wasn't implemented in bulldozer. But I guess there are some things that we just will never know.
Well after reading through a few reviews the verdict is out.
The bad:
-Improvement over zambezi in single threaded tests was only 8-12%
-Improvement over zambezi in multithreaded tests was only 12-16%
-No improvement in power consumption over zambezi
-Almost all of the performance improvement over zambezi comes from a clock rate increase (which they were able to do thanks to switching the design over to the same high density logic gate library that their GPUs use) IPC has only increased by 2-6% over zambezi.
-Power scaling with overclocking has only slightly improved over zambezi. It still exhibits huge increases in power consumption from slight overclocks and requires voltage bumps even for minor overclocks. 10-20% OC is the highest you can possibly go on most high end air/liquid cooling setups and that results in another 15% increase in performance. That slight overclock will increase the chips power consumption by more than 50%.
-Intels processors continue to beat the living tard out of it in single threaded tests
-Intels processors continue to beat it by a reasonably significant amount in mixed tests
-Intels processors continue to beat it by a small amount in multithreaded floating point heavy tests
-Intel processors continue to beat the living tard out of it in games (I actually don't fully understand why, what kind of code profile do video games usually fall into?)
-Intels processors continue to use significantly less power both at idle and at full load
-Intels processors continue to offer better performance per watt on average
-Intels high end processors come with an IGP while AMDs do not
-Intels processors continue to OC by higher margins with less voltage increase and with much less power consumption/heat increase
The good
-The clock rate increases did not increase power consumption. Therefore stock power consumption levels are still tolerable.
-It is now without question better than phenom II in all applications not just most.
-They are priced surprisingly low.
-The FX-8350 goes head to head with the 3570K and 3770K in multithreaded integer heavy tests, which most slow applications these days are. It beats the 3570K far more often than it loses in this area. Similar situations can be found in the other models. They consistently beat higher priced sandy/ivy bridge processors in multithreaded integer applications.
-Therefore performance per dollar can be better or much better than sandy/ivy bridge at stock depending on what type of applications you're going to be running.
-We all know that the architectures performance is subpar with certain types of applications. But the gap between AMD and Intel is narrower now for those applications than it was before.
Conclusion: It's still not great, but it competes with Intels low end and medium end offerings fairly well and unlikely zambezi you could actually justify this purchase in some scenarios. It can reach the performance of the 3570K and 3770K in many applications for less money but requires more power to do it and doesn't have as much overclocking headroom available. It's exactly what we all expected as far as the level of improvement goes. What is impressive is that the performance improvement going from zambezi to vishera is greater than the performance improvement going from sandy bridge to ivy bridge was. And Intel had the advantage of moving to 22nm while AMD achieved that greater performance improvement on the same manufacturing process. What doesn't make sense is why AMD didn't use their high density library to begin with, considering most of the performance improvement came from that and this library has been around for years I don't see why it wasn't implemented in bulldozer. But I guess there are some things that we just will never know.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson
"I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. "
-Mark Antony
-Ron Swanson
"I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. "
-Mark Antony