AMD Athlon 64 FX-60: A Dual-Core farewell to Socket-939
by Anand Lal Shimpi on January 9, 2006 11:59 PM EST- Posted in
- CPUs
Final Words
The prospect of dual core usually meant that you had to give up some single-threaded performance, but with the Athlon 64 FX-60, that tradeoff is no longer true. While it's still not clocked as high as the fastest single core processor, it is extremely close. It's close enough that the difference in clock speed makes single-threaded performance a non-issue.
The fact that the FX-60 and FX-57 are almost at a clock speed parity is important because it does mean that we are getting beyond the inflection point of the dual core/single core transition. The high end dual core chips are all but caught up to the high end single core chips of just six months ago. Intel still has a little way to go before the same is true for them, but with the FX-60, it does hold true for AMD.
Unfortunately, since the FX-60 is still built on the same 90nm Toledo core as the previous X2 processors, overclocking headroom is not that great. With a retail AMD heatsink/fan, the best that we could do is 2.8GHz at 1.40V. With more exotic cooling, you could probably manage better, but stepping up the voltage all the way up to 1.50V wouldn't yield a 3GHz overclock on air.
Then there's the issue of AMD's upcoming Socket-AM2; due out in another few months, you obviously won't be able to use any Socket-939 processors in the new motherboards and there will be no upgrade path beyond the FX-60 for current 939 owners. So, our recommendation would be to stay away from the FX-60 unless you absolutely have to build the world's fastest system today.
If the latter is true, then you can't go wrong with the FX-60; if not, however, you'll be better off waiting for AM2.
The prospect of dual core usually meant that you had to give up some single-threaded performance, but with the Athlon 64 FX-60, that tradeoff is no longer true. While it's still not clocked as high as the fastest single core processor, it is extremely close. It's close enough that the difference in clock speed makes single-threaded performance a non-issue.
The fact that the FX-60 and FX-57 are almost at a clock speed parity is important because it does mean that we are getting beyond the inflection point of the dual core/single core transition. The high end dual core chips are all but caught up to the high end single core chips of just six months ago. Intel still has a little way to go before the same is true for them, but with the FX-60, it does hold true for AMD.
Unfortunately, since the FX-60 is still built on the same 90nm Toledo core as the previous X2 processors, overclocking headroom is not that great. With a retail AMD heatsink/fan, the best that we could do is 2.8GHz at 1.40V. With more exotic cooling, you could probably manage better, but stepping up the voltage all the way up to 1.50V wouldn't yield a 3GHz overclock on air.
Then there's the issue of AMD's upcoming Socket-AM2; due out in another few months, you obviously won't be able to use any Socket-939 processors in the new motherboards and there will be no upgrade path beyond the FX-60 for current 939 owners. So, our recommendation would be to stay away from the FX-60 unless you absolutely have to build the world's fastest system today.
If the latter is true, then you can't go wrong with the FX-60; if not, however, you'll be better off waiting for AM2.
Facebook
Twitter
RSS
94 Comments
View All Comments
shortylickens - Tuesday, January 10, 2006 - link
Is it just me or does AMD seem to be moving architectures along a little too quickly?Socket A had an incredibly long lifespan. It moved from Athlon B's to C's (thunderbird) to XP to MP and even a Duron thrown in for good measure.
With the tow of the latest passing on (754,940) it seems like AMD just cant get it right. Now they are are already peaking with socket 939. From now on, we'll only have pity chips thrown to us just so we dont feel bad about going into it.
Well I feel bad. :(
Nyati13 - Tuesday, January 10, 2006 - link
Socket A was around for a long time, but there were 4 different FSB specs, several different chipsets, and 2 different RAM types all lumped under Socket A. You couldn't tell for certain that a Socket A CPU would work with any random Socket A motherboard because of all the spec changes. You can say that any Socket 939 CPU will work in any Socket 939 motherboard.DigitalDivine - Tuesday, January 10, 2006 - link
That's the disadvantage of having an integrated memory controller. Memory types change, and in the past there have been numerous memory types, in order to take advantage of the changes, amd has to change their socket.jakerugged - Tuesday, January 10, 2006 - link
In the review under the gaming tests you say - "We should also mention that we had to re-run our AMD numbers in this test since the last review as we were seeing sub-par AMD performance. A clean install and re-run of the numbers yielded the results you see today; the Intel numbers didn't change." Why did the Intel numbers NOT change and why did the AMD numbers change? Was it only in the Gaming tests or once you had good AMD numbers in the gaming tests did you then run the SYSMARK, etc tests in order to get max performance? If you did thats not good testing methodolgy. Can we see these sub-par numbers, how bad are they?Its ok to say that this CPU is fast but only if you add in that you may have to repeatedly re-load and re set-up your system until it is "just so" before it gives the kind of performance that you would expect.
Betwon - Tuesday, January 10, 2006 - link
About the of FX-60, they(both the websites) find:"our processor worked just fine at 2.8GHz clock speed. Unfortunately, Athlon 64 FX-60 didn’t get along with the 15x clock multiplier."--from xbitlabs
"With a retail AMD heatsink/fan, the best we could do is 2.8GHz at 1.40V. With more exotic cooling you could probably manage better, but stepping up the voltage all the way up to 1.50V wouldn't yield a 3GHz overclock on air."--from anand
They agree.
jakerugged - Wednesday, January 11, 2006 - link
Whoa there.....This has gone on a little bit longer than even I thought.My original comment was about the general stability of the FX60, because Anandtech.com just put in a little one liner about how they had to re-load the entire FX60 test rig (Im assuming this means re-install the O/S, all drivers, patches and benchmark tools.) because it was not performing correctly. They still have not said why or what these low figures were?
To be honest, Im not really into O/Cing but I am into stability and I dont want to have to re-load my system just so I can play a few games after using Word or sending a couple of e-mails.
Betwon - Tuesday, January 10, 2006 - link
About the of FX-60, they(both the websites) find:"our processor worked just fine at 2.8GHz clock speed. Unfortunately, Athlon 64 FX-60 didn’t get along with the 15x clock multiplier."--from xbitlabs
"With a retail AMD heatsink/fan, the best we could do is 2.8GHz at 1.40V. With more exotic cooling you could probably manage better, but stepping up the voltage all the way up to 1.50V wouldn't yield a 3GHz overclock on air."--from anand
They agree.
PrinceGaz - Tuesday, January 10, 2006 - link
Why don't you quote a bit more of what xbit labs said, Betwon?"We managed to get our CPU to work stably with the clock frequency multiplier set to 14x without increasing the Vcore, which equaled 1.3V throughout the entire test. In other words, our processor worked just fine at 2.8GHz clock speed. Unfortunately, Athlon 64 FX-60 didn’t get along with the 15x clock multiplier.
Having increased the Vcore by 10% above the nominal, we still couldn’t get our hero to run stably at 3GHz frequency. Although the CPU would boot the Windows XP just fine and could even go through some test applications, it would still crash to the Blue Screen of Death (BSOD) when both cores were fully loaded. So, we had to give up or desperate intention to conquer the 3GHz height.
We managed to get our processor to work absolutely impeccably at 2.9GHz with the Vcore set to 1.44V. This frequency was achieved as 14 x 207MHz."
So they managed 2.9GHz totally stable, not just 2.8GHz. And on the final page
"However, we shouldn’t say that Presler got completely and hopelessly defeated by the new Athlon 64 FX-60. Due to the new finer production technology, the new dual-core processors from Intel can boast excellent overclocking potential. As for the frequency potential of the AMD processors, it has been almost completely exhausted by now. As a result, when we compare the results demonstrated by the overclocked AMD Athlon 64 FX-60 and Intel Pentium Extreme Edition 955, the former is not always the winner. So, if you do not mind your system being a little bit noisy and generating quite a bit of heat, then Pentium Extreme Edition 955 might be a way to go."
Note the wording-- "the former [o/c FX-60] is not always the winner". The meaning being that the FX-60 at 2.9GHz beats the EE955 at 4.26GHz more often than not, but there are situations where the o/c EE955 wins. So when both are overclocked to their max, they are either equal or the FX-60 has a slight advantage overall. Of course the FX-60 has a significant advantage when it comes to heat, power consumption, and noise.
Betwon - Tuesday, January 10, 2006 - link
I had say that:"OC:
955 is 4.26GHz, FX-60 is 2.9GHz -->Now, the ratio is 1.46 (4.26/2.9). -->P4 starts to be better than X2. "
Note:
"As for the frequency potential of the AMD processors, it has been almost completely exhausted by now."
"Due to the new finer production technology, the new dual-core processors from Intel can boast excellent overclocking potential."
flyck - Tuesday, January 10, 2006 - link
but they also benchmark them both... and p4 is overclocked in those benchmarks still slower then X2. And they also say the power consumption of that overclocked p4 is insane.