Amazing Bang-for-Buck? AMD Athlon II X4 631

[LOL_Wut_Axel]

Member mosox posted in another thread this legit Chinese review of the Athlon II X4 631. This is a socket FM1 Llano CPU, which means it has its IGP disabled. It also means that the reference clocks aren't as integrated as the APUs, so overclocking should be easier and better than what we've typically seen from Llano. With that in mind, here's what they were able to achieve overclocking on stock voltage on an MSI A75MA-G55:

32nm Llano. 4.43GHz on stock voltage=decent power consumption. Quad-Core. ~10% higher IPC than Zambezi (Bulldozer). Cheap motherboards. No IGP. $90 for the CPU.

Looks like it scales very well with higher frequencies. For a cheap rig this looks like an absolute winner, easily a better choice than the FX-4170 and if you're aiming for high performance at a cheap cost also much better than the Core i3-2100. Only downsides I think would be the upgrade path if desktop Piledriver isn't socket FM1 compatible, and the IGP is disabled so you have to use dedicated. If you're overclocking you're likely to have one, though.

I think this one steals the cake from the Phenom II X4 955 when it comes to bang-for-buck.

 

[StrangerGuy]

That should show AMD how to do a 32nm chip right...

Whoops.

 

[Maximilian]

More like this :thumbsup::thumbsup::thumbsup:

 

[alyarb]

Only downsides I think would be the upgrade path if desktop Piledriver isn't socket FM1 compatible.

If trinity is based on piledriver and if piledriver is anything like bulldozer there's a chance this will be the only 32nm gimped APU that is any good, particularly on power, and that is the downside because unless trinity is some spectacular product you'll only using the board for this one chip. I'd like to see more benches and more samples, though.

As far as "best AMD value" it'll come down to either this or the 960T.

 

[Broheim]

I was actually thinking about getting a secondary rig with one of these a couple of days ago, throw on a NH-D14 (I have a spare) OC its pants off and run it to hell and back. I'm just waiting to see other people's experience OCing these things.

 

[podspi]

It's likely you'll have an "upgrade" path to Trinity w/ FM1 (although it isn't officially confirmed, AFAIK). I wonder how many of these chips can actually hit that speed, though. If they all can (on stock volts), you have to wonder why AMD would over-volt them as much as they did.

My guess is that you'll get some great chips that can do this, and some that won't do anything. Luck of the draw.

 

[BlueBlazer]

AMD should be using this and add another 4 cores, instead of Bulldozer. :thumbsup:

 

[antisocialmunky]

Lol, I made a post about this 2 days ago then 2 people post this today. Funny stuff.

 

[Thermalzeal]

As I said somewhere else. This SKU, and it's existence, clearly pinpoint the GFoundries issue is with the graphics part of llano core. I read they had 60% yields, I wonder how much of the 40% failure is this SKU.

 

[antisocialmunky]

Lots but its an Athlon that lives up to being an Athlon.

 

[TakeNoPrisoners]

For reference my X4 955 gets 10.623 sec on Wprime 32M at 4.0Ghz.

 

[-Slacker-]

4.4ghz from 2.6 ghz on STOCK VOLTAGE????

I uh ... what?

 

[mosox]

Member mosox posted in another thread this legit Chinese review of the Athlon II X4 631. This is a socket FM1 Llano CPU, which means it has its IGP disabled. It also means that the reference clocks aren't as integrated as the APUs, so overclocking should be easier and better than what we've typically seen from Llano. With that in mind, here's what they were able to achieve overclocking on stock voltage on an MSI A75MA-G55:

Glad we can shed some light on the capabilities of some obscure AMD CPUs that doesn't seem to interest many people.

I had two reviews for that CPU I only posted the newer one in my thread because I already posted the other one in here:

http://forums.anandtech.com/showthread.php?p=32336835&highlight=#post32336835

Their OC was much smaller though that's why I would like to see a reputed site like Anand do a review.

Also is there any reason the FM1 can't be turned into a nice budget gaming platform? Take a look at this mobo:

http://www.newegg.com/Product/Produc...82E16813131784

 

[antisocialmunky]

Llanos are already nice budget gaming systems but this would propel them into the realm of 2100 and 2400 killer.

I'd like to see a write up on what 'Lab Burst Mode' actually does though.

 

[Squirtmonster]

No level 3 cache?

 

[Vesku]

Curious as to what the teased Llano Black Editions will show. AMD furiously working to add some L3 cache?

 

[LOL_Wut_Axel]

No level 3 cache?

Not gonna happen. Too many transistors need to power it, increasing power consumption significantly. To overcome the cache deficiency these have 4x1024KB instead of 4x512KB, or 4MB L2 vs 2MB L2. They also have higher IPC by about 4%, so yeah.

Curious as to what the teased Llano Black Editions will show. AMD furiously working to add some L3 cache?

Looks like the A8-3870BE would be too expensive at ~$160, even taking into account the "free" $50 graphics card included in the die.

AMD will not release these with L3 cache, mostly because it's unneeded because of what I mentioned above.

 

[Vesku]

AMD will not release these with L3 cache, mostly because it's unneeded because of what I mentioned above.

Not seeing anything in your original post, or are you just referring to clocking past what L3 would gain? Would have to see what power draw is like but I think from a performance per watt perspective that L3 would beat out squeezing more MHz.

 

[RobertPters77]

Wow. I had one. Maybe I should've done a suicide run with it instead of giving it away.

 

[LOL_Wut_Axel]

Not seeing anything in your original post, or are you just referring to clocking past what L3 would gain? Would have to see what power draw is like but I think from a performance per watt perspective that L3 would beat out squeezing more MHz.

What I wrote above:

Not gonna happen. Too many transistors need to power it, increasing power consumption significantly. To overcome the cache deficiency these have 4x1024KB instead of 4x512KB, or 4MB L2 vs 2MB L2. They also have higher IPC by about 4%, so yeah.

It would increase the die size and decrease yields.

 

[SickBeast]

These chips are interesting for sure. Hopefully we'll see some cheap bundle deals for them. At 4.4ghz, they would make great gaming CPUs.

 

[Kevmanw430]

Imagine what one of these could do on water cooloing with a voltage bump?

 

[Silverforce11]

AMD messed up big time not releasing LLano with a BE edition. Being on 32nm its got a lot of headroom for OC.

 

[SickBeast]

Imagine what one of these could do on water cooloing with a voltage bump?

We don't even know if it's stable at 4.4ghz. Even if it is, it will be barely faster than a Phenom II X4 at 3.8ghz, which many people have had for over a year now, myself included.

I don't think I would feel comfortable putting more than 1.4v into a 32nm CPU. Perhaps there is more headroom, but I doubt it.

Not only that, but you'll have to deal with a company that lies to its customer base and releases a plethora of sockets unnecessarily.

 

[LOL_Wut_Axel]

We don't even know if it's stable at 4.4ghz. Even if it is, it will be barely faster than a Phenom II X4 at 3.8ghz, which many people have had for over a year now, myself included.

I don't think I would feel comfortable putting more than 1.4v into a 32nm CPU. Perhaps there is more headroom, but I doubt it.

Not only that, but you'll have to deal with a company that lies to its customer base and releases a plethora of sockets unnecessarily.

I don't think I'd call ~17-18% faster "barely faster".

Also, you can't compare directly the CPU voltage of AMD and Intel on the same process when it comes to degradation. AMD CPUs are quite a bit more tolerant of voltage, and this should do 1.45V just fine for three or more years if it has a good cooler to keep temps low.