Preproduction Socket AM2 CPU Test

Omega65

THG gets their hands on a engineering sample AM2 X2 4800+ and compares it's performance to the current S939 4800+.

View: A Look At AMD's Socket AM2 Platform

As our benchmarks show, going from DDR400 to DDR2667 with current silicon would not make a noticeable difference at all, because the integrated memory controller suffers more from relaxed memory timings than it can gain from speeding up clock speed via DDR2. Whether the current engineering sample processors actually suffer from a memory controller bug is hard to say. This could also be due to information that was released intentionally to prevent people from doing early benchmarking.

I didn't expect DDR2 (533, 667) to add any performance. The jury is out on DDR2 800 though

Source: Toms Hardware

RWB

I wouldn't figure you'd see any kind of performance increase. I mean your compairing a 4800+ to a 4800+

Gobbles

RWB wrote:

I wouldn't figure you'd see any kind of performance increase. I mean your compairing a 4800+ to a 4800+

yes but your also comparing ddr to dd2 and as noted the timing have to be so lax on the ddr2 that it nulls performance gains on the lower end memory so that the only benefit you may see is with the ddr2 800 stuff.

Ray

it makes you think if the graduates who design these memory modules really payed attention in class or did all the math....

lemonlime

Yikes, Not terribly impressive I'm afraid. I was happy to see that they mentioned official support for DDR2-800. They are going to need super high frequencies like that to make up for the crappy timings. As of right now, I see little reason to move away from 939 unless there are other memory controller tweaks (for better handling of 2GB and multiple dimms for example..) not to mention resolution of the cold bug, although I wont hold my breath on that one..

As we already mentioned in our evaluation of ULi's M1697 chipset, there is no need for new chipsets for AM2, and it won't affect the chipset development of 3rd party manufacturers either. Basically, every core logic product that supports the current AMD64 processor family will also support the Socket AM2 generation, because the communication path between the processor and the system is HyperTransport.

Way to go AMD .. someone give these guys a 100% pure 1337 award for their A64 architecture already!

I have totally lost track of the number of chipsets and boards that intel has gone through.. who knows what chip works with what chipset

Omega65

Remember that this is a Pre Revision F Sample. Rev F supposedly has an updated and revised DDR2 Memory Controller.

Also there wont be any more speed bumps for S939. The 2.6ghz X2 5000+ and 2.8ghz FX-62 will only be on AM2.

lemonlime

Omega65 wrote:

Remember that this is a Pre Revision F Sample. Rev F supposedly has an updated and revised DDR2 Memory Controller.

Also there wont be any more speed bumps for S939. The 2.6ghz X2 5000+ and 2.8ghz FX-62 will only be on AM2.

Thats a good point.. I guess its only a matter of time before the enthusiast crowd moved to AM2 for the higher end chips.

a2jfreak

What's the "cold bug?"

lemonlime wrote:

I see little reason to move away from 939 unless there are other memory controller tweaks (for better handling of 2GB and multiple dimms for example..) not to mention resolution of the cold bug, although I wont hold my breath on that one..

Omega65

Usually when you cool a CPU to subzero temps you get a (much) better Overclock. CPUs that have the "Cold bug" wont even run when at Subzero.