I thought it was stable with memtest but I was wrong, once I got into windows it started acting up all crazy. Although I did manage to get a screen dump before it took a memory dump
Please be aware the score is likely affected by the slower overclock of 4.4GHz.
And I have this as well, I pushed up the FSB so I could change the memory as close as possible, as I figured it can't be to far away from the stable limit if it can do somewhat tight timings at 1600MHz therefore I brought the memory speed up to 1800MHz and the core clock to 4.5GHz
I thought it was stable but I was wrong, once I got into windows it started acting up all crazy. Although I did manage to get a screen dump before it took a memory dump
Please be aware the score is likely affected by the slower overclock of 4.4GHz.
If you could get those timings to be stable then I would say its a "slight" bit better than 1600@7-8-7-20, same reason I couldn't do it is my 2133 ram wouldn't rock tight timings at 1866 stable either.
Today is finally the right time to part II, test with AMD Vishera FX
INTRODUCTION After a long time I finally got to the being playing with the new processor Vishera FX. Generally processor rather exceeded my expectations. His performance in practice is everywhere higher than the Zambezi FX and also improved the potential for clocking and of course due to the power consumption of the MHZ clock. I'm quite pleased with the power consumption after considerable overclocking, which is actually quite a bit lower.
Roughly speaking, the new FX processor at the same clock offers a 3-25% higher performance than the previous generation of FX processor. In the most common case, the differences of about 10-15%. That's not bad for a simply architecture refresh on the same production process. The processor has been in the eye only these tiny changes and I think that has been streamlined many processes (bugs removed) from the previous concept. Even this, however, counts, and so Vishera Bulldozer is already more capable the bulldozer to the masses.
But now let's take the motherboard Asus Crosshair V Formula-Z, a description of you may have read two months ago, it is in principle perhaps the very best piece of hardware processors Bulldozer for Vishera I uploaded already newer BIOS, otherwise the set is fully identical.
I tested on these programs. Unfortunately, for reasons of time, I got more demanding, but all the practical tests, we were able to see some world-class reviews ( example www.hardware.fr ). Even so, I have included several graphs to compare and Vishera Zambezi FX to their default values and then after maximum "stable" overclocking. Individual tests include benchmarks: Memory subtests: Aida64, MaxxMem Calculation of pi: very oldschool superpi 16M, newer wprime 32M a 64M, older PiFast aka Hexus, modern MaxxPI a System stability tester used two methods by focusing on one core. 3Dmarky a games test: 3D06, 3Dvantage, 3D11, DosBox, Unigine Heaven DX11 Rendering: Cinebench R10, Cinebench R11.5, POV-Ray 3.7 Video: x264 4.0, x264 FHD CPU complex test: Fritzchess benchmark, CPU 3Dmark Vantage, Physics 3Dmark 11, Geekbench, UCbench Working with files compression and decompression: Winrar 4.20, 7-zip
Graphs So first joint charts, where both FXs side by side...
3Dmark 11 –purely score for the processor
Memory subtests in AIDA64, more MB/s is better
Rendering in POV-Ray
Elderly x87 SuperPi (interesting that I came to you the result of default better for Vishera, though SuperPi is Vishera processor for some unknown reason, or worse your efficiencyu)
The greater part, however, graphs associates Vishera look at the results on nominal settings against a maximum overclocking.
You have full 3DMark tests and results
Memory test MaxxMem in bandwith GB/s
Scores in Cinebench R10 for one and all working and computational benchmark kernels Nuclearus.
A také novější Cinebench R11.5
Complex test Geekbench a Fritzchess
SuperPi may be counted as using other alternatives. System Stability Tester can do it through one or more cores and with different way ... MaxxPi loaded with all cores and utilizes the newer instructions, there was a meaningful calculation of the 8M places.
Wprime loads all cores, PiFast was not ...
Ucbench used to calculate both brute force and in podvýpočtu then some newer SSE instructions
DosBox uses very low resolution benchmark Quake1 is thus dependent on brute force single core CPU without much optimization. Unigine Heaven running on the contrary, FHD resolution with DX11.
Winrar and 7-zip indicate the strength of compression and decompression. This is after OC, especially with 7-zip, really huge.
X264 in the second passage is representative videoencoding
Overclocking is a chapter that offers most of the articles also often differs term "stability" as well as the methodology of determining stability. I think it is not strictly necessary to worry processor 10h in the most brutal tests of stability. Stability is for me to state computer assembly from the perspective of the user, in any application where there are no crashes, restarts and OS freezing after a period of use. For some it may be constantly running Folding @ Home, for another 3h PRIME95 and for some stability in online games ... I tested the 30min AMD Overdrive stability test. From my perspective, this is a very Uptake test, where you can select different parts of load that we want to test or classically tick everything and have a complete test stability
The alpha and omega of AMD processor temperature. Still I have to repeat over and over and temperature, temperature, temperature. If we follow Coretemp, FX cores temperature should not exceed the long term value of 70 ° C (this is in reality the CPU still around 85!). The following, however, we keep the temperature the better. Example. The default settings and low overclocking we can afford to keep the temperature and 70C will see without instability. But if we on the maximum of the CPU clock, there us already this temperature can limit. 4.9 GHz and a temperature of 70 C does not bode well for a while everything will be okay, gradually we will see probably error. However, if at the same clock and voltage reaches a temperature knock to 60 C, I'm sure that the processor will be fully stable. It must be really good cooling - high quality cooler with pushl-pull fan (the bigger the flow, the better) and sprite or mainstream allinone like Corsair H70 etc.
Because of this I got to the beautiful 5010 MHz! It could indeed take over, and 4950 MHz to put the clock tuning, where temperatures are not as much of a problem. It is a higher average, but on the internet I've seen a few better.
The overclocking is also related small chapter I love above all, and the maximum overclock for benchmarks. Here I got an additional 200 MHz up Cinebench I even went through with efforts to 5240 MHz! Below is a short video jerky quality (probably nerves as I was tight ) and below some pictures.
Undervolting at stock clocks
Not as we see chapter of undervolting. And with the attendant decrease in consumption. For many, it is more interesting than the actual overclocking. And also because he devoted an integral part of the text and images. As you can see in the first picture in the load voltage declined in CPU-Z at 1212 V.
Another image is already the completed 30-minute state of stability, I opened it AI Suite II to see realistically set voltage to the load, the idle voltage shown in CPU-Z (but I do not know why the new CPU-Z does not appear here podtaktování cores ...) . What is even more interesting and beneficial of undervolting is sometimes higher performance! This is due to the fact that the frequency of the processor in the base varies between 4000-4200 MHz and is subject to TDP of CPU. The lower voltage consumption is obviously reduced and thus the load is often a multi-threaded processor 4100 MHz frequency.
And how everything is reflected in consumption? Meant, what is the consumption in the base, and when overclockingor undervolting ground state? Well.
Among states in idled are not significant differences. Approximately 3W savings even worth mentioning. Basically, the entire assembly during exercise tells about 232 W. After overclocking over the 5 GHz consumption increased, but the processor is not as uncomfortable hungry as FX Zambezi. Consumption grew by "only" 83W! This is about more than 70W less than the FX-8150 at a frequency of about 4880 MHz with the same line-up! If the processor to operate default values of frequencies, we are able to save with undervolting decent 36W. With longer tweaking of CPU voltage VDD and CPUNB would be still some Watts knocked down. All, of course, also depends on the length of the processor.
Crosshair combo with Vishera FX-8300 can be recommended to all fans who are looking for maximum platform at AMD. The board will delight stability, widgets (fastboot, diretckey, post display, etc.), software, build quality and components fitted, fully digital cascade (DRAM and CPU), processor FX-8350 offers very good indeed multithreaded performance and respectable single thread, finally overcomes him with confidence Phenom II (mainly in practical use and gaming). Is definitely a huge plus overclock with good cooling and definitely my plus earn significantly lower power consumption for overclocking than the last bulldozer.
That's a difference of a little under .4%. I would say that's pretty damn consistent. When I would not change priority on my old i7 920, I would usually see it bounce between 6.6 and 6.5, which is a 1.5% difference.
It didn't make any difference at all for me. I get 8.31 consistently.
Originally Posted by Deedot78
Add me please.
Im using RASA XPC 360. Temp are fine, chip just wont budge anymore.
A chip that's worse than mine, woohoo!!
Originally Posted by Deedot78
Thanks for Tip. Much better now, it was stumping me.
Got Vcore down to 1.41 for 4.6.
Thanks @ Flank3r, very nice info and tests. Thanks for also confirming my estimates I made a while back.
Originally Posted by ComputerRestore
I'm going to estimate that a good starting baseline with these chips, would be
- 4Ghz @ 1.2v (maybe even less)
- It would be nice to see it at 4.8Ghz @ 1.35v and under 50c on a decent Air Cooler. My 8150 that was 30% less effecient could do 4.8Ghz at 1.44v
Going to guess they would top out around 5.5Ghz under water @ around 1.55v
Also by reading a lot of other overclocks on these forums, I'm starting to wonder....
a) Low Voltage overclocks on these chips seem almost entirely Chipset Based (Cooler running chipsets require much less CPU voltage)
b) Chipset temperatures over a certain threshhold cause a need for really high VCore during overclocks.
c) How effective would a Chipset Waterblock be on PileDriver? (6Ghz Overclclock stable?)
This is all a guess, but it might be worth looking into. Putting a fan over the chipset may not be enough either, if the heat transfer to the Heatsink isn't great.
I'd recommend for anyone having high VCore issues to repaste the Chipset Heatsinks.
It's also worth mentioning that they are not always installed properly as well. For example
- Crosshair IV Formula and Exteme both had an issue where they used the wrong size stand-off under the heat sink. The Heat Sink, barely touched the North Bridge Chip - and would overheat during normal operation, let alone Overclocking. Edited by ComputerRestore - 11/13/12 at 6:41am
Run this one by me again, in English this time.
You can not chose to "boost the single thread". Performance gain is performance gain, what it effects is everything.
my bad yes it effects everything rather I was speaking of comparison between thuban and buldozer/piledriver where thubans loved higher NB (more or less on mobo) what i was trying to get at is seems like the limitation is on chip and not on board
as i said i am seeing a lot of single core boost with the multi rather than fsb and ram, in short i am asking if we know what portion of the architecture is the bottle neck.. if its the prefetcher or something else