Richland - the best OC APU - infos, experience, issues

[alpha0ne]

Thanks for the review FlanK3r

Its getting close to retiring my old 2500K/POS Anus board and a new A10 is looking like the next replacement

[The Stilt]

The Book of Bulldozer - Revelations: Chapter 1 (Intro)

Something is about to change within couple of days...

Meanwhile, my first proper run on a A10-6800K.
It is an old (1303SUS) part and a weaker specimen in general (atleast on LN2).
The leakage is simply not high enough, I am running into the voltage limit ("saturation").

On a better chip 8.2GHz+ should be quite doable.



http://valid.canardpc.com/show_oc.php?id=2828933

[FlanK3r]

wow, thats awesome!

[Mechanical Man]

The Book of Bulldozer - Revelations: Chapter 1 (Intro)

Something is about to change within couple of days...

...[/URL]

I hope this with 5GHz Vishera was not what you meant.

[The Stilt]

I hope this with 5GHz Vishera was not what you meant.

Nope.
It is something for the existing products, Zambesi, Vishera, Trinity & Richland.

[FlanK3r]

Il try Richland with Ln2 in next 10 days . Stilt, this was your "5600 MHz" air chip?

[The Stilt]

Yes it was.

Flank3r, when you'll see "it" you'll 5h1t bricks.
Still, please do not tattoo my name on your bum

[Lightman]

Yes it was.

Flank3r, when you'll see "it" you'll 5h1t bricks.
Still, please do not tattoo my name on your bum

Haha!

I bet it's new version of Turbo Core called Stilt Turbo! It's so good you can use it only once, after that there is no return

Can't wait for reveal!

[tbone8ty]

TurboStilt

[Mechanical Man]

Stilt

Couple days have been passed, and no bricks from flanker yet!.

[The Stilt]

http://www.xtremesystems.org/forums/...SuperPI-x87%29

Can I have my bricks now?

[FlanK3r]

Mechanical man: Im not busy with dark side (HSWL) for mobo review, back to AMD 22.6. ,-)

[The Stilt]

Did some undervoltage testing on a A10-6800K.
I have modified the motherboard to allow the telemetry reading from the VRM controller, however the readings seem to be inaccurate due bad calibration. Thats the reason I only used clamp meter this time.

The readings are for the whole system (everything that is connected to the PSU, excluding the single SSD drive).
The VRM losses are included in the readings.

The PSU voltages remained at exactly the same levels during all of the tests.

12V = 12.130V
5V = 5.02V
3.3V = 3.375V

Setup:

AMD A10-6800K 4.1GHz / 1.5-1.8GHz (NB) / 844MHz (GPU)
ASUS F2A85-V PRO (VRM at default settings, 50% CPU & NB Load-line, phase shedding enabled)
AMD Radeon Gamer Series 16GB DDR-2133 DRAM (4x 4GB Hynix MFR - Single Rank)
AMD FX closed loop watercooler (fans at 40% DC)

Cool & Quiet, Turbo Core (CPB) & CC6 = disabled.

These are the absolute peak values, the average power consumption is significantly lower.

1st run (default settings):

CPU Core VDD = 1.3125V
UNB Core VDD = 1.2750V (during 3D stress)
DRAM VDD = 1.6500V
DRAM Timings = CL10-11-11-27-2T

Idle:

12V rail combined = 2.24A (27.171W)
5V rail combined = 0.94A (4.718W)
3.3V rail combined = 1.28A (4.32W)

Total power = 36.21W

X264 R2334 10-bit:

12V rail combined = 6.83A (82.847W)
5V rail combined = 1.44A (7.228W)
3.3V rail combined = 1.28A (4.32W)

Total power = 94.39W

Note: Preset = Slow, Tune = Film, CRF = 21.0

Prime95 V27.9 LargeFFT:

12V rail combined = 8.36A (101.406W)
5V rail combined = 2.25A (11.295W)
3.3V rail combined = 1.29A (4.353W)

Total power = 117.055W

FurMark Burn-In:

12V rail combined = 7.25A (87.942W)
5V rail combined = 3.67A (18.423W)
3.3V rail combined = 1.29A (4.353W)

Total power = 110.719W

Tomb Raider Benchmark:

12V rail combined = 6.63A (80.421W)
5V rail combined = 2.80A (14.056W)
3.3V rail combined = 1.34A (4.522W)

Total power = 99.000W

Hitman Absolution Benchmark:

12V rail combined = 7.87A (95.463W)
5V rail combined = 3.43A (17.218W)
3.3V rail combined = 1.33A (4.488W)

Total power = 117.170W

Resident Evil 6 Benchmark:

12V rail combined = 8.05A (97.646W)
5V rail combined = 3.43A (17.218W)
3.3V rail combined = 1.34A (4.522W)

Total power = 119.387W

WCS - Prime95 LargeFFT / FurMark Burn-In / HCI-Memtest:

12V rail combined = 10.80A (131.004W)
5V rail combined = 3.72A (18.674W)
3.3V rail combined = 1.32A (4.455W)

Total power = 154.133W

2nd run (hand tuned voltages, 24/7 stable):

CPU Core VDD = 1.2250V
UNB Core VDD = 1.1500V (during 3D stress)
DRAM VDD = 1.3500V
DRAM Timings = CL12-11-11-27-2T

Idle:

12V rail combined = 2.07A (25.109W)
5V rail combined = 0.74A (3.714W)
3.3V rail combined = 1.31A (4.421W)

Total power = 33.245W

X264 R2334 10-bit:

12V rail combined = 5.62A (68.170W)
5V rail combined = 1.09A (5.471W)
3.3V rail combined = 1.32A (4.455W)

Total power = 78.097W

Note: Preset = Slow, Tune = Film, CRF = 21.0

Prime95 V27.9 LargeFFT:

12V rail combined = 7.03A (85.273W)
5V rail combined = 1.66A (8.333W)
3.3V rail combined = 1.33A (4.488W)

Total power = 98.095W

FurMark Burn-In:

12V rail combined = 5.83A (70.717W)
5V rail combined = 2.61A (13.102W)
3.3V rail combined = 1.35A (4.556W)

Total power = 88.376W

Tomb Raider Benchmark:

12V rail combined = 5.11A (61.984W)
5V rail combined = 2.01A (10.090W)
3.3V rail combined = 1.37A (4.623W)

Total power = 76.698W

Hitman Absolution Benchmark:

12V rail combined = 6.41A (77.753W)
5V rail combined = 2.42A (12.148W)
3.3V rail combined = 1.34A (4.522W)

Total power = 94.424W

Resident Evil 6 Benchmark:

12V rail combined = 6.68A (81.028W)
5V rail combined = 2.20A (11.044W)
3.3V rail combined = 1.33A (4.488W)

Total power = 96.561W

WCS - Prime95 LargeFFT / FurMark Burn-In / HCI-Memtest:

12V rail combined = 8.71A (105.652W)
5V rail combined = 2.58A (12.951W)
3.3V rail combined = 1.34A (4.522W)

Total power = 123.126W

Results (default configuration vs. hand tuned):

Idle: Peak Power = -8.92%
X264: Peak Power = -20.87%
Prime95: Peak Power = -19.33%
Furmark: Peak Power = -25.28%
Tomb Raider: Peak Power = -29.08%
Hitman Absolution: Peak Power = -24.09%
Resident Evil 6: Peak Power = -23.64%
WCS: Peak Power = -25.18%

Average: Peak Power = -22.04%

Throw in a Radeon HD 7750 and I'd say thats a <200W well spent

HD 8670D + HD 7750 in Dual Graphics is:

100%+ faster than a HD 8670D alone (Tomb Raider)
33%+ faster than a HD 7750 alone (Tomb Raider)

[FlanK3r]

Second 6800K is here...now only more time for testing,. but I have not yet (making review of Z87 boards+new FX)

[Calathea]

Guys, do richland oc better than trinity? Since I have a very hard time to sell msi FM2 atx mobo I'm considering buying either 750k or 760k for gaming together with GTX580. Sadly I need an old trinity cpu to update bios so I'm thinking if it's worth the hassle of finding a bios flashing cpu just to get 760k later or if I should buy the 750k instead.

[The Stilt]

Sadly I need an old trinity cpu to update bios

No you don't.
Richland will post just fine on any FM2 motherboard.

[Evantaur]

only have FM1 board, would like to buy richland but no FM2+ mobos and FM2 mobos will be useless when kaveri comes... :P

[Miwo]

Waiting for FM2+ as well... id like to see more Asym crossfire options than just the 6570/6670/7570/7670/7750. Cards are getting cheap enough than it just doesnt make sense to go buy a lower end video card for the sake of Richland/Trinity crossfire.

Just last week they were selling a 7770 for $75 with no rebates and free shipping. At that price, it makes no sense to get a 7750, pair it with a Richland, and still not outperform the 7770. Asym with ALL GCN Radeons is key to make this a more attractive platform for people other than OEMs and HTPCs

[Calathea]

No you don't.
Richland will post just fine on any FM2 motherboard.

lol wtf just ordered a trinity for bios flash. Call me stupid, cause there's no other explanation why I have not tested richland on the mobo yet. Daymn.

Alright. Tried it out. Posted fine and flashed bios on ASUS F2A85-M PRO. Only black screen on MSI FM2-A85XA-G65.... ASUS - MSI = 1 -0

[The Stilt]

The retail Richland parts are simply superior compared to the old sample parts:



http://valid.canardpc.com/show_oc.php?id=2871078



This is the first retail part I've ever tested.
As you can see, the screenshot and the validation show different cores.
Thats because both of the compute units on this chip can do 8.2GHz

[FlanK3r]

wow, batch number? Looking very good 8200+

[The Stilt]

wow, batch number? Looking very good 8200+

1319PGN.

This chip has base leakage of 76.923%.
The older 8GHz part had 53.846% leakage, which is too low because Richland does not scale properly past >2V VDDCR.
I got another part from the same batch with 69.230% leakage. Need to check it later today.

However the leakage alone simply does not define the overclockability of a part.
The ultra high leakage parts usually do worse than the parts with medium/low leakage properties no matter the cooling used.

The rules also seem to be different for Vishera's too.
The highest clocking parts (on all cooling methods) seem to be medium / low leakage parts, while the high and ultra high leakers seem to run like crap even on LN2.

On Richland around 8.4G should be possible on a optimal specimen.
I am certain that the chip I tried today is able to do atleast 8.3G with some further tuning.

[The Stilt]

The second 1319 part can also hit similar frequencies.
7.8GHz required just 1.77V.

The chip also sat on the desktop at 8060MHz with 1.9V for couple of minutes without any issues.
It requires slighty higher voltage than the previous one, however the voltage scaling appears to be better.

[FlanK3r]

very nice, so you are ready for 32M superpi? From my side, 12-16.8. Il be at LN2

[leo5111]

can anyone tell me what to set and what to turn off to OC? i will be useing a 6800k a asus http://www.newegg.com/Product/Produc...82E16813131883 ASUS F2A85-M PRO FM2 AMD A85X (Hudson D4) ram is gskill sniper ddr3 2133 and i will be useing the onboard video thanks