-
Killer Trinity - top A10-5800k - confirmed 5.1GHz on air and 7.3GHz on LN2
-
thank you :D
dat voltage...
-
Very nice! Now please show us some 2D benches to get an idea how fast Trinity really is. ;)
-
And 3D too (no NDA at 3D) ;)
-
A bit high voltage for 7.3GHz indeed ;)
-
thank you guys ;)
Quote:
Originally Posted by
FlanK3r
And 3D too (no NDA at 3D) ;)
tonite i will run some 3d bench ;)
Quote:
Originally Posted by
The Stilt
A bit high voltage for 7.3GHz indeed ;)
yes ;(
under 4500MHz they run very well but all my chips requires voltage to perform at higher clocks..
here some 2D action
http://downloads.criminalcafe.com/re...aida%20cpu.jpg
http://downloads.criminalcafe.com/re...aida%20fpu.jpg
http://downloads.criminalcafe.com/re...res/aida64.jpg
http://downloads.criminalcafe.com/re.../cinebench.jpg
http://downloads.criminalcafe.com/re.../pi%207zip.jpg
http://downloads.criminalcafe.com/re.../pov%20ray.jpg
-
Good job ! Someone with a 4150 clocked@4 ghz can run the same benches so that we see the piledriver improvements?
-
1 Attachment(s)
Attachment 130381
Quote:
Originally Posted by
Alex-Ro
Good job ! Someone with a 4150 clocked@4 ghz can run the same benches so that we see the piledriver improvements?
I am running the benches on my Phenom II to see how it compares.
http://img.techpowerup.org/120930/Capture042040.jpg
http://img.techpowerup.org/120930/Capture043.jpg
http://img.techpowerup.org/120930/Capture044.jpg
-
how bout combined overclock of cpu and gpu?
can yours do 4.5ghz cpu with a 1ghz+ gpu? on air?
how bout high speed memory? got a 2400mhz kit?
-
Povray is 20% faster than on FX4100,C11.5 is 13% faster than on FX4100. I suspect that 5800K run at 3.8Ghz in MT benchmarks (not at 4Ghz as CPUz reports).
-
Quote:
Originally Posted by
informal
Povray is 20% faster than on FX4100,C11.5 is 13% faster than on FX4100. I suspect that 5800K run at 3.8Ghz in MT benchmarks (not at 4Ghz as CPUz reports).
You really think that CPU-Z is reading the multiplier wrong by exactly 2x?
-
It's reading the (full core) Turbo clock since Trinity will almost always try to run at max clock possible. Default is 3.8Ghz though and this is the clock it will run at if you push a MTed workload. Maybe in winrar which is integer heavy it will run at full core turbo(4Ghz) since FP unit is not taxed. You can test this if you still have FX. Just run C11.5 and watch the clock in real time. FX8150 won't run above 3.6Ghz. Same goes for povray and similar fp intensive tests.
Of course,pure ST benchmarks will always be run at 4.2Ghz ,no matter what is the type of workload.
-
Unless there is a significant amount of GPU processing power required, all of the cores of A10-5800K will run at 4.0GHz during Cinebench or similar NORMAL (such as X264 encoding) multithreaded software. During execution of a software which does not fit in the definition of "normal use" (e.g. Prime95) the frequency may temporarily drop down to 3.8GHz or even lower. However only in cases where the power envelope / thermal entity reserved for the CPU is exceeded.
If GPU processing power is required simultaneously, the Bi-directional Apm feature will balance the CPU and GPU clocks to have the best possible performance for both of them.
-
Quote:
Originally Posted by
informal
It's reading the (full core) Turbo clock since Trinity will almost always try to run at max clock possible. Default is 3.8Ghz though and this is the clock it will run at if you push a MTed workload. Maybe in winrar which is integer heavy it will run at full core turbo(4Ghz) since FP unit is not taxed. You can test this if you still have FX. Just run C11.5 and watch the clock in real time. FX8150 won't run above 3.6Ghz. Same goes for povray and similar fp intensive tests.
Of course,pure ST benchmarks will always be run at 4.2Ghz ,no matter what is the type of workload.
Thanks, I didn't realize what you meant there.
It is possible that AMD has improved Turbo further on Piledriver though, so I wouldn't make comparisons with Turbo enabled in MT workloads just yet.
The jump in turbo's effectiveness (performance gain related to time enabled) went from .5-1% on Thuban (never worked) to about half of intel's solution on Bulldozer/Llano so I'm interested to see what Trinity does.
Quote:
Originally Posted by
The Stilt
Unless there is a significant amount of GPU processing power required, all of the cores of A10-5800K will run at 4.0GHz during Cinebench or similar NORMAL (such as X264 encoding) multithreaded software. During execution of a software which does not fit in the definition of "normal use" (e.g. Prime95) the frequency may temporarily drop down to 3.8GHz or even lower. However only in cases where the power envelope / thermal entity reserved for the CPU is exceeded.
If GPU processing power is required simultaneously, the Bi-directional Apm feature will balance the CPU and GPU clocks to have the best possible performance for both of them.
Are you quite sure?
Cinema 4D / Cinebench is quite a workload, while I understand that Cinebench is a rather short load it is a 100% load that I have found can heat the CPU up just as much as Prime95 when looped.
-
I think that Stilt might very well be correct here since Trinity is able to split the power budget . I'm still not so sure that povray and C11.5 will constantly run at 4Ghz during whole runtime but I think it is quite possible.
-
1 Attachment(s)
I'll throw a locked llano into the benchmark pile. VM folding stable for 3 days (24h/day) now at current settings.
I know the a10 is stock clocks but I'm not going to put mine back, so this is what you get. Open gl is with low nb/ram since I'm folding and higher volts/clocks just waste power.
Attachment 130382
-
Quote:
Originally Posted by
tbone8ty
how bout combined overclock of cpu and gpu?
can yours do 4.5ghz cpu with a 1ghz+ gpu? on air?
how bout high speed memory? got a 2400mhz kit?
very easy ;)
i can run 3d11 at 4700/1100MHz (cpu/gpu) on air
and 4800Mhz is OK for cinebench
i still working on memory clock ,2133MHz is OK
but i still tweaking
Quote:
Originally Posted by
informal
Povray is 20% faster than on FX4100,C11.5 is 13% faster than on FX4100. I suspect that 5800K run at 3.8Ghz in MT benchmarks (not at 4Ghz as CPUz reports).
Informal,
its 4GHz,the turbo is off
i chose 40x100MHz on setup for a start ;)
-
Thanks for the info PcCI2iminal!
So 4.8Ghz is doable for C11.5? Nice to hear. I suppose real stability is some 200Mhz below that on air? It's pretty good all around product I must say :)
-
I'm afraid that CPU wise Trinity might not be able to pull ahead from Llano on a consistent basis...
Looks like CPU overclocking is a little worse than Zambezi based CPUs but not too bad overall for what it is.
I'm excited for these. I want to see affordable mini-itx boards from ASRock, Gigabyte and ASUS :) Should make great mini-HTPCs/media boxes. I'd like to throw a build together in one of those cheap HEC Mini-ITX cases. HEC PSUs are usually decent enough to where I wouldn't worry.
@informal
Real stability on FX was about 400 less than Cinebench but if heat output of Trinity is extremely manageable then maybe that can be closed up some?
-
Re-tested the clocks on two different chips.
Please note that the parts I am using might not fully represent the ones available to public.
The chip I first tested seems to constantly hold at Pb1 PState (4.0GHz) during Cinebench / X264 encoding.
After around 10 minutes of X264 encoding, the second chip will start to back down to PS0 (3.8GHz) about half of the time, making the average frequency 3.9GHz. This chip has a higher voltage rating for Pb1 PState so the power draw is a bit higher than on the first one.
The PState registers were monitored with a 200ms sampling rate and the settings used for X264 were the heaviest ones:
X264 x64 10-bit (R2216)
Input content: "Raw" Avatar 1080p m2ts (Blu Ray)
Settings: Single pass, Preset: Placebo, Tune: Film, Bitrate 9000, Output format: mkv.
Luckily Trinity can usually be heavily undervoltaged ;)
-
PcCI2iminal:
Can you please check which components the Biostar board uses, for: External Pll (Realtek xxx) and CPU Vrm controller?
-
@ BeepBeep2
Yeah you are probably correct. Real stability will be ~4.4Ghz most likely but this is not bad at all. Sure not on the level of FX4170 or upcoming FX43xx but still pretty close I guess.
@ The Stilt
Single pass means 1st pass ? Can you test with 2nd pass setting? IIRC the 2nd pass is much more taxing on the multicore CPUs.
-
Quote:
Originally Posted by
informal
@ The Stilt
Single pass means 1st pass ? Can you test with 2nd pass setting? IIRC the 2nd pass is much more taxing on the multicore CPUs.
During the 1st pass in 2nd pass encoding mode the CPU load is very low.
In a single pass encoding mode the CPU load is at the maximum level whole time.
-
Quote:
Originally Posted by
PcCI2iminal
very easy ;)
i can run 3d11 at 4700/1100MHz (cpu/gpu) on air
and 4800Mhz is OK for cinebench
i still working on memory clock ,2133MHz is OK
but i still tweaking
Informal,
its 4GHz,the turbo is off
i chose 40x100MHz on setup for a start ;)
Nice job,however regarding memory clocking biostar boards are TERRIBLE,i just sold my biostar z77x hi-fi,unable to boot anything higher than 2133 with same cpu and ram that would do 2600+ on a maximus 5 gene.I hope we see some 2600+ ram on trinity pretty fast,but that won't happen on biostar i'm afraid...
-
Quote:
Originally Posted by
Alex-Ro
Nice job,however regarding memory clocking biostar boards are TERRIBLE,i just sold my biostar z77x hi-fi,unable to boot anything higher than 2133 with same cpu and ram that would do 2600+ on a maximus 5 gene.I hope we see some 2600+ ram on trinity pretty fast,but that won't happen on biostar i'm afraid...
Trinity unofficially supports DDR-2400 DRAM ratio.
Currently it is available on some of the Asus and Gigabyte motherboards atleast.
