All posts related to V2
User avatar
By polynurb
#357663
concerning the oc-issues i agree with RenderFred, there are some entries in there that are misleading.
and the non-geeks can't spot these easily.
to put +4ghz on any dual xeon in production is nonserious...

also i am not sure if the core-counts are always correct, concerning Hyperthreading.
eg "Genuine Intel R CPU 2.90GHz" i think is a dual cpu machine (hpz620), maybe with HT turned Off?

still i think overclocking results are very interesting when they are tagged as such and give sufficient info.
(ram type, fsb settings, vcore/dram voltage, prime stable??)
i guess the problem is to read out the actual clock speed automatically&accurately.
maybe this could be included in the future? (maybe something like cpu-z grabbing infos silently?)
..because it seems people are likely to skip the "additional info", doesn't matter then if intentionally or due to sloppiness.

daniel
By photomg1
#357672
Just did a little test using the old bench scene to get a better understanding on the network overhead. I must say maxwell scales incredibly well on my couple of machines.

machine 1 bench of 976.84 , machine 2 bench of 2210.84 combining those two would be a bench of 3187.68. The actual network render came in at 3118.93.I shaved off 25 percent of my overall time by doing the network render.

Anyway I do agree if you have overclocked your cpu ,you should post that info(cpu speed ,is it stable etc.) in your cpu bench results.
By feynman
#357684
...and we good and ever so respectful designers that take this week's Rio+20 very seriously, should actually be undercklocking - or go back to manual gouache rendering ;)
User avatar
By abgrafx3d
#357767
I'm trying to run the benchwell scene on v2.7 and I get this error "Cannot download benchwell data."

Nevermind - got it working after I logged in to the mxm gallery.
Last edited by abgrafx3d on Mon Jun 25, 2012 4:02 pm, edited 1 time in total.
By l1407
#357771
photomg1 wrote:Just did a little test using the old bench scene to get a better understanding on the network overhead. I must say maxwell scales incredibly well on my couple of machines.

machine 1 bench of 976.84 , machine 2 bench of 2210.84 combining those two would be a bench of 3187.68. The actual network render came in at 3118.93.I shaved off 25 percent of my overall time by doing the network render.

Anyway I do agree if you have overclocked your cpu ,you should post that info(cpu speed ,is it stable etc.) in your cpu bench results.
oc of 2600k and 3770k are very stable 8)
By numerobis
#357772
l1407 wrote:
oc of 2600k and 3770k are very stable 8)
an oced system has to be completely stable (not VERY) otherwise it's useless...
By Ha_Loe
#357777
numerobis wrote:
l1407 wrote:
oc of 2600k and 3770k are very stable 8)
an oced system has to be completely stable (not VERY) otherwise it's useless...
Well, there isn't any completely stable sytem, or is there? It's about probability of failure. If you run your CPU at default clock with the boxed fan, it's going to heat up to x degrees and last y hours with a probablity of 99.99%.

As long as you don't hit any frequencies that mess up the internal timing, failure typically is due to excess heat or insufficient power supply. CPU ratings have to hold for very cheap cooling (boxed fans) or cheap power supply. If you can make sure, that you have plenty of stable power (good power supply and mainboard power regulators) and good cooling (typical OC fans or water or whatever you like.), then there is no reason, the OCed system should be any less stable than the default CPU with cheap hardware.

OCed setups either fail quickly due to messed up timings or are essentially as stable, as the default clock speed.

I have my 2600k running at 4.3Ghz. I could push to about 4.6 without even adjusting the voltages. I've had this rig render for days and run 24h for weeks without any problem.

Anyway, I can't see why you should post an unstable config for benchwell.
By numerobis
#357793
Ha_Loe wrote: Well, there isn't any completely stable sytem, or is there? It's about probability of failure. If you run your CPU at default clock with the boxed fan, it's going to heat up to x degrees and last y hours with a probablity of 99.99%.
Yes... sure. I only wanted to say that an overclocked system (if it is done right) doesn't have to be less stable than a system at stock speed, like this "very" implicated. ;)
User avatar
By AndreD
#357798
numerobis wrote:
Ha_Loe wrote: Well, there isn't any completely stable sytem, or is there? It's about probability of failure. If you run your CPU at default clock with the boxed fan, it's going to heat up to x degrees and last y hours with a probablity of 99.99%.
Yes... sure. I only wanted to say that an overclocked system (if it is done right) doesn't have to be less stable than a system at stock speed, like this "very" implicated. ;)
++ ;-)
By feynman
#357827
numerobis wrote:
Ha_Loe wrote: Well, there isn't any completely stable sytem, or is there? It's about probability of failure. If you run your CPU at default clock with the boxed fan, it's going to heat up to x degrees and last y hours with a probablity of 99.99%.
Yes... sure. I only wanted to say that an overclocked system (if it is done right) doesn't have to be less stable than a system at stock speed, like this "very" implicated. ;)
+1

Today, motherboard vendor's automatic oc features are ever so easy to use and stable in running, provided no extravagant hw components are used in one's build. Every studio I know runs their render hw oc'ed with no fuss.
User avatar
By gianca
#357847
polynurb wrote:concerning the oc-issues i agree with RenderFred, there are some entries in there that are misleading.
and the non-geeks can't spot these easily.
to put +4ghz on any dual xeon in production is nonserious...
<snip>
still i think overclocking results are very interesting when they are tagged as such and give sufficient info.
(ram type, fsb settings, vcore/dram voltage, prime stable??)
i guess the problem is to read out the actual clock speed automatically&accurately.
maybe this could be included in the future? (maybe something like cpu-z grabbing infos silently?)
..because it seems people are likely to skip the "additional info", doesn't matter then if intentionally or due to sloppiness.

daniel
"to put +4ghz on any dual xeon in production is nonserious"?
FYI Boxx has been selling dual x5690 OC to 4.5Ghz for a while. I think they are still selling them.
I'm guessing they are likely using a EVGA SR-3 mobo + dual liquid cooling.

About my system (Gianca 3960X)...
I wasn't sure about how readable will be the benchwell info panel if would put too much text, so I opted to just write the very basic info.

For people interested in knowing more, here's it:

Intel i7 3960X
ASUS Rampage iV Extreme
32Gb 1600 Memory
System Disk Seagate Velociraptor 600Gb 10k rpm

I'm running this system 24/7 for CG production, using 3 OC settings: 4.7Ghz, 4.8Ghz and 4.9Ghz.
They are all stable settings, but I only tested up to 4,8Ghz overnight: I haven't the luxury to test my 4.9Ghz setting yet for overnight stability, but with that settings under %100 load it did not go over 76C.
I started building my own OC systems a couple years ago with a 980X, and I've been using them in production since and I only had a dozen blue screens, mostly at the beginning with experimental settings.
The only reason why I'm on 4.7 instead of 4.8 is that my room is currently missing AC, so it's not worth to run the risk even if seems to stay under 80C.

I know that an OC cpu will not last long, but in my business I only need these for a couple of years before moving on to something faster.

My 0.2 cents.

Gianca
By feynman
#357848
gianca wrote:I know that an OC cpu will not last long, but in my business I only need these for a couple of years before moving on to something faster.
That's the point; an Intel i7-3960X is 841,00 € - just about a senior designer's daily rate - so if it does fail one day: just plonk a new one in and continue working...
User avatar
By Fernando Tella
#357851
feynman wrote:That's the point; an Intel i7-3960X is 841,00 € - just about a senior designer's daily rate
I wanna work where you do.

Here is getting near a senior designer's monthly rate (just a bit more). No joke.
User avatar
By Hervé
#357872
heehe.. looking at salary; Germany and Spain are two very different countries..

h/

Thanks Miguel I know 5.2+ is a big development hi[…]

SketchUp 2021 Update

Hi Alberto, Many thanks for taking the time to ma[…]

Let's talk about Maxwell 5.2

Features are closed and tested, we are testing p[…]

Materials Library not working

Yeah. We have changed to a new website on a new se[…]