Cosmos 2 Build log - [In Progress] ...continued

[nicej]

na man i just increased multiplier. even left memory clock on auto

 

[nicej]

I will look into that QbobWatson. Thanks mate

 

[Modz 4 Higher]

A couple things. Nicej, since you told Mort you wanted as detailed as possible explanation get prepared to read a novel He goes way into detail so just be prepared.

Prime does not stress as much as IBT and LinX. They say on IBT max level you only need to complete 5 passes. I always do 20 but be ready for your temps to go through the roof. Also try using Furmark for stressing

 

[QBobWatson]

Prime does not stress as much as IBT and LinX. They say on IBT max level you only need to complete 5 passes. I always do 20 but be ready for your temps to go through the roof. Also try using Furmark for stressing

IBT and LinX are functionally the same, right? They're both just frontends to Linpack?

 

[Modz 4 Higher]

Yes the both use identical Linpack Libraries just have a different GUI

 

[nicej]

Thanks for the headsup guys. Yea my temps were pretty high I thought during my IBT max runs. I'm going to go to bed now and leave IBT max running all night. I'm just going to set like 300 runs and see where it's upto when I get up.

Here is my 3D Mark 11 P-Score:
http://3dmark.com/3dm11/3076907;jsessionid=1jl1a6edc7x3w18jbvwk021kq4

going to try pushing the GPU further later on.

 

[nicej]

Here is a benchmark of my RAID0 Setup of the two HyperX's using ASSSD and ATTO:

Not even kidding.

 

[nicej]

Here's a SuperPi test run at 1M:

 

[nicej]

I only did 3 runs of max stress, to get on with some benchmarking. In all honesty that is "stable enough for me" and I will be running max stress IBT all night when i hit the sack anyway.

Here's a screen shot of the 3runs and some temps. I thought it was running pretty hot, what do you think?

 

[nicej]

Check out this pic of my station, from old pc to new one LOL

That's a 24" monitor btw...It's just dwarfed next to the Cosmos II.

 

[Modz 4 Higher]

Looks good. I was like.... Going to bed????? It's noon time over here. Forgot you were in Austrailia.

 

[nicej]

yea lol. If im going to be honest.. im a little disappointed in my 7970. I'm not gonna lie - I was expecting more.

I run bf3 in everything ultra including highest possible antialiasing I get about AVG 80fps imo, which I was expecting to be higher since its factory OC'd

 

[nicej]

im wondering if i should leave my OC as it is at 4.9, or try to go higher by tweaking those skews and paying around with it.

i should be happy with 4.9 though right? i mean it's not like its slow Lol

 

[Modz 4 Higher]

I'd leave it at 4.9. Hell the stock clock is plenty fast. Lol

 

[nicej]

Here are results of a prolonged MAX stress IBT run, that ran while I was napping. It turns out it had finished its 12th run, and I decided to stop it and turn it off to get some rest

201 - 202 GFlops. Get on my level!


Here's a shot of the temperatures, I didn't remember to scroll down HWmonitor so just check out the max temps:

 

[Modz 4 Higher]

Looks good man

 

[m0r7if3r]

as detailed as possible without the need of years of experience to understand. detailed enough to be educational! haha

OK, I'm going to assume a basic understanding of the fundamentals of semiconductor technology here, if you lack that, wikipedia article for transistors or something like that will be more than sufficient to get you up to speed. The way the manufacturing process for semiconductors works is rather simple. You take a semiconductor (typically silicon, but germanium is starting to see a rise in popularity in some of the hsld stuff like the new solar panels and such...OT but interesting) and you create, often through a chemical reaction, an oxidized layer on top of it. In the case of silicon, you're looking at SiO². This is your starting point for anything based on semiconductors.

From here, to create a transistor, you have to remove some of the oxide. The means by which this is done is with really powerful lights and such through a process called etching (which is really complicated and not relevant enough ot the discussion to warrant explanation, sufficient to say that the oxide is removed). The important thing to note here is that the whole wafer is not done at once, but rather individually from the same mask through movements of the mask and/or the wafer to get correct placement of the pattern on the wafer. This process is repeated many times to create a multi-layered pattern (with an oxide removal step added).

The wafer is then tested as a whole, then split to chips, then those chips are tested and such...the actual process from here is generally irrelevant, as the differences in the silicon have already been established. For those of you following along at home, now is your time to solve the mystery! Where was the difference in the chips introduced? That's right! It has multiple sources! The one you probably didn't think of was impurities in the silicon. Without a thorough discussion of material properties, it's difficult to explain how this could affect the clocks the processor could attain, but the long and short of it is that it changes the conductivity of the material (which actually affects many other things......but that's a different discussion). The other one comes from the etching process itself. The machines that do the etching can only be so accurate, and while they are some of the most accurate ever seen in the history of this planet, the variations are minor enough that different chips can have different properties. This can come from anything from the mask being a little skewed to the lighting having been uneven. This is also why sometimes a certain chip out of every batch will clock better, the chip in the position that gets the most perfect etching will clock the best.

It should be noted that I've summed up a lot of stuff here...this is by no means comprehensive, I tried to keep it simple. There are other sources of error, but those are your main 2. Also, take all of this with a grain of salt, I'm pretty tired right now, so I might have messed something up, if you catch anything, please call me out on it!

 

[Modz 4 Higher]

I told you he would right a novel. Now ask him to be as technical as he can He would need his own thread

 

[nicej]

That is actually really easy to understand. I should have asked for more details! But yea i read all that and enjoyed reading it.

What do u guys think about my benchmarks scores?

 

[Modz 4 Higher]

Glad you understand. Maybe he will be nice enough to go more in depth for us *hint* *hint*