ASUS Crossblade Ranger Reviewed - FM2 OC monster in need of a new APU

[Monstru]

AMD?s line of mainstream APUs has been on the market for some while, but ASUS didn?t think they were in need of a ROG motherboard. All through the FM1 (Llano) and FM2 (Trinity, Richland) users all over the world expressed their disappointment to ASUS about not launching a ROG motherboard for these platforms. The wait is over now because ASUS just launched its first ROG motheboard targeting AMD?s mainstream APU line: Crossblade Ranger. Now we just need a new APU that can take advantage of this new FM2 monster OC board...

ASUS Crossblade Ranger Review

[FlanK3r]

nice readings Monstru!
There is my: http://www.xtremesystems.org/forums/...usiast-warrior

[Monstru]

Nice stuff!

[PatRaceTin]

Good to see Asus ROG in FM2 line.

[Lightman]

Nice board and very nice reviews!
I agree that this board begs for new APU! There is one coming up but I don't think it will OC that good. I hope I'm wrong here

FM2+ is a nice platform for HTPC or causal gaming as I'm using it exactly for that with my A10-7850K, but CPU itself can't be pushed clock wise as much as 32nm predecessors could.

[paulbagz]

Even if you use two top end GPUs, won't the CPU hold them back?

-PB

[ilkkahy]

I think it would be nice to see amd apu with six cpu cores and bit fewer gpu cores. It would have somewhat decent cpu power and hsa features.

[tbone8ty]

Where are OC results? Im gonna say this board doesnt offer to much more performance?

[FlanK3r]

look to the my review, auto OC results are IN from yesterday. Now Im doing manual, hope tomorow will be complete.

[Monstru]

Where are OC results? Im gonna say this board doesnt offer to much more performance?

The current APU's don't offer too much performance, that is the problem...

[FlanK3r]

my OC is from 3.7 GHz(4GHz turbo) up to 4.5 GHz stable. But I know about some Kaveri 4.65 GHz full stable . x264FHD, 32M, ycruncher, wprime1024 or CInebenchs and of course 3D Marks iGPU are not bad with Kaveri. 4.5 GHz Kaveri is better than 5 GHz Richland in average.

[The Stilt]

There is really no point in overclocking Kaveri.
The desktop Kaveri "high-end" parts are already pretty much maxed out at stock and for that reason the power efficiency is quite horrible.

The "critical frequency" for current Kaveri silicon is around 4.2GHz and that range can be achieved at stock voltage (or even lower) levels and with any kind of cooling.
On 28nm BULK the operating temperature of the chip doesn't seem to matter much, if at all. All of the domains clock the same no matter if entry level air cooling or a custom water loop is used, as long as it doesn't throttle because of the excess temperatures.

If you insist of going higher than 4.2GHz you really need to start pushing the voltage.
Usually getting 4.4GHz stable needs >0.1V additional voltage over the base voltage.
4.5GHz or even higher might be doable, but not at safe voltage levels.

Besides being unsafe it is also completely pointless as the power draw increases so heavily.
The increase in power consumption is far greater than the increase in performance.

E.G. (A10-7850K):

3.7GHz / 1.2875V = ~66W CPU Pmax
4.2GHz / 1.2875V = ~70W CPU Pmax
4.4GHz / 1.3875V = >84W CPU Pmax
4.5GHz / 1.4750V = >101W CPU Pmax

The maximum achievable frequency will always be limited by the voltage.
On Kaveri silicon permanent damage starts to occur at 1.49V.

[FlanK3r]

for light benchs I using 1.55V, friend, how many Kaveri chips did you tested? Your example si as my average 7850K , exactly with the voltage

[The Stilt]

Honestly, way too many.
The behavior is the same no matter what model it is.

The production parts are admittingly slightly better than the prototype or ES parts, but just slightly.
The general behavior is identical as it is a manufacturing process / design limitation (similar to the 4.6GHz limit on Orochi).

Vishera has way better performance to watt ratio compared to Kaveri (at frequencies beyond Kaveri's real capabilities = >3GHz) but it's overall performance is also significantly better, thanks to the 4 CUs and L3.

Eventhou the 32nm SOI process was heavily bashed in the beginning, no one expected that things could go even worse.
The 32nm SOI is superior in every way but in density compared to 28nm BULK.

And in 28nm BULK it will remain for quite a while...

[FlanK3r]

hm, with refresh of Kaveri seems too