Maxwell Render | Official public forum

http://reviews.zdnet.co.uk/hardware/des ... 700,00.htm

* * *

deadalvs

added: HELL YES !

* * *

deadalvs

and how many licenses do you need to play with Maxwell on that?

two... NL didn't change that policy... yet !

* * *

deadalvs

from what i saw on alienware.com the octocore config is a bit cheaper than a quad config

netherless i should mention that the speed gain is only of 1.375
the four core config with 5160 xeon top at 1600 cinebench while the octocore tops at 2100-2200 ( maybe not enough cache and small bus)

so overclocking couldn't help either when the bus bandwidth (1.3 GHZ) is the bottleneck ?

* * *

deadalvs

i dont think so, but i'm not a mobo expert

Is the Cinebench test most applicable to rendering? Because that molecular modeling benchmark they ran is just about twice as fast.

sure that the scene used in the cinebench is quite and old one that doesnt even use gi and take seconds to render, maybe they should update it to one more adequate to current standar ( gi Ao whatsoever )

don't they use radiosity ??

it samples always these red points and then renders the image... cinema always used this to render "GI"... or am i wrong...

cinema people?

* * *

but cinebench is at least good because it gives a feedback for the overall system speed.

* * *

maybe it could be an advantage based on maxwell's computive intense equations so that the bus isn't fully used while rendering... just assuming, i have no idea ...

i HATE still having bottlenecks others than on wine or beer in these times... !

* * *

deadalvs

deadalvs wrote:so overclocking couldn't help either when the bus bandwidth (1.3 GHZ) is the bottleneck ?

* * *

deadalvs

I think overclocking will still help. Depends also what renderer you're testing, I think Maxwell will benefit a lot from the octocore system, perhaps even close to 100%, while other renderers that use buckets will see a little less improvement.

Mihai wrote:

deadalvs wrote:so overclocking couldn't help either when the bus bandwidth (1.3 GHZ) is the bottleneck ?

* * *

deadalvs

I think overclocking will still help. Depends also what renderer you're testing, I think Maxwell will benefit a lot from the octocore system, perhaps even close to 100%, while other renderers that use buckets will see a little less improvement.

can You go a little deeper - this sounds VERY interesting...

i'm planning to invest, You know... !

* * *

we should guess how an 8 core system on win64 with 4 gigs of ram would perform with the maxwell1.1 benchmark... i think it would take around 23 minutes.
more ? less ?

* * *

deadalvs

it is a 2.66 quad.
the 4x 2.66 has 1400
the 8x2.66 has 2100

so the speedgain is 1.5 not 1.3 over a quad system!
i think thats not to bad

cheers
stefan

yes the overall is not bad at all but we must take into account the price it would cost in term of rendering licence maybe two mob in a core 2 quadro would give more for the buck, dont know... at least i think with the current data available that we should wait a bit for nexts octosystems that will have less cache bottleneck...anyway still the kinda system i would take if it was offered to me

why do You think there's a specific bottleneck based on the cache ? each processor has 8 MB L2 cache...

or which cache... ? L3?

which one is important for renderings and why ?

* * *

deadalvs