Read more.Low-power options down to 25 watts.
Read more.Low-power options down to 25 watts.
Wonder if we'll see any of those coming to an AM3+ socket any time ... a 45W 2.8GHz piledriver with up to 3.8GHz turbo is pretty tasty...
Argh, bloody AMD - their website says these are C32 but their quick-reference document says they're AM3+ ... wonder which is true...
Yeah, 4300 being C32 makes sense - given the 2 HT links it can't really be AM3+ But you can get 1P on AM3+ and C32, so the 3300 could be on either.
Hopefully some of these will trickle into the FX lineup...
The Opteron 3200 series was AM3+ so it makes sense the Opteron 3300 series is AM3+ too.
Another thing though.
The earlier 3250HE,had 4MB of L3 cache:
http://www.amd.com/us/products/serve...cessors.aspx#5
This indicates half the modules were deactivated.
Unless it is a typo,the 3350HE supposedly has 8MB of L3 cache:
http://www.amd.com/us/products/serve...cessors.aspx#4
This would indicate one core per module is now deactivated.
It should mean IPC is up over a normal PD quad core.
Most interesting product release in a long time from AMD IMO. I can see a lot of the 3300 range drifting into enthusiast quiet/fanless home user machines if they release an AM3+ version. Certainly got me interested.
I thought L3 cache amounts were independent of core count? After all, there were Phenom II X4's with reduced L3 cache. If L2 cache amounts were different I could see where you were coming from, but I'm pretty sure this is just a case of them having the power envelope to run more cache (AFAIK cache memory actually consumes a reasonable amount of power). Presumably this is down to the increased efficiency of Piledriver*. The 3280 - 3380 comparison shows that the two processors are identical apart from a 200MHz step up in clock speed. Presumably with the from the 3250 the 3350 has bumped clocks 300MHz, but AMD had already released the 3260 at 45W, and the 3350 only has a 100MHz bump over that. I'd wager that, for the 3300 quad-cores, they chose to enable the full L3 cache rather than push the clocks up - presumably this is better for server workloads....
*EDIT: or, it just occurs to me, it could be down to Piledriver having better yields on the now-mature 32nm process and so there being more dies with a full 8MB L3 that will function at lower voltages...
There are currently 1 users browsing this thread. (0 members and 1 guests)