[deadalvs]

i need a faster machine ...

* * *

so it seems every xeon calculation we did in earlier threads were not even close to correct.

[mashium123]

Wow! Great!

17 minutes... man...

[Thomas An.]

I am happy to report that all is right in the world today.

Now ! .... we are talking good.
(Previously it felt weird that the octacore benched lower than the quad.)

[Frances]

Yay! Congrats Chris.

[ivox3]

..thanks all.

Mihai: it is ?? cool, ...had no idea. lol

Thomas: ..yeah,.. I think we all felt somewhat disturbed by that -- I definitely did.

So folks,.. let me ask a question, ... what do you think we could expect with some 2.67's and maybe more ram ? That's not going to happen anytime soon over here,...but now curiosity is truly peaked.

[michaelplogue]

Ceeee-riminy!

Now, in the chart, we need to have a price per benchmark ratio......... Take the total cost for the computer at the time of purchase, and divide it by the benchmark number. The smaller the figure, the better "bang for the buck" (BFB) value....

[-Adrian]

So good to hear that your investment is finally worth it for maxwell, a near linear performance increase.

[Mihai]

Now, in the chart, we need to have a price per benchmark ratio......... Take the total cost for the computer at the time of purchase, and divide it by the benchmark number. The smaller the figure, the better "bang for the buck" (BFB) value....

Yes, good idea. Although with prices changing so often you would need to constantly update it. It's nice to see in the case of ivox 8 core, an almost 100% scaling.

ivox, so you now have two sticks, both running at 667mhz?

[ivox3]

Yes.

..and the S5000XVN MB can do a total of 32GB.

@ Adrian : ...thanks thanks, .. me too.

[glebe digital]

that's warp speed and no mistake.

[michaelplogue]

OK.. I have a question. I haven't run the test scene recently, but I can't imagine that it commits a huge amount of memory. So (in theory) as long as you had enough free RAM to cover the commit charge of a single instance of MXCL - regardless of how many cores you have - I wouldn't expect to see any difference by adding more RAM. Or am i wrong and each core commits equal amounts of memory?

I would have thought that 2GB of RAM should have been more that sufficient in rendering the scene without having to access VM on the hard drive. Why would adding more RAM have had such an increase in speed?

[Thomas An.]

OK.. I have a question. I haven't run the test scene recently, but I can't imagine that it commits a huge amount of memory. So (in theory) as long as you had enough free RAM to cover the commit charge of a single instance of MXCL - regardless of how many cores you have - I wouldn't expect to see any difference by adding more RAM. Or am i wrong and each core commits equal amounts of memory?

I would have thought that 2GB of RAM should have been more that sufficient in rendering the scene without having to access VM on the hard drive. Why would adding more RAM have had such an increase in speed?

I think it has to do with balancing of the memory sticks. They have to be in pairs (both slot 1 and 3 for example) so that the CPU can function optimally .. but that was on AMD systems, so I am not 100% sure on this.

[ivox3]

Michael, ... yes -- I was having similar thoughts. So what I'm doing now is re-running an array of tests. It'll be 3 tests each with the CPU priority set to normal, above normal and high. I'll average the results I get into a singular score for each setting.

* Also, ..I'll be taking actual time readings with a digital stop watch for the actual render times.

It turns out that the CPU's were set to high on that last BM from a previous test and I forgot about it. When it's setup like that the Maxwell clock doesn't update properly, ..so that time may be faulty.

So, .. I ask that everyone disregard that last score for now until I have some more accurate testing to leverage it with.

[Prowler]

Hi ivox3,

I really hope the 17 minutes are not a bug related to Maxwell's time measuring!

Would you mind doing also the latest Cinebench on your machine? Would be great!

Here is the link:

http://www.maxon.net/pages/download/cinebench_e.html


Thanks!

[ivox3]

I'll run the Cinnebench a little later ...