FX-8350 CPU/DIE Orientation + Raystorm Mod

[Johan45]

I found that the CHV-z and Sabertooth were both very capable boards but the CHV-z seemed to be a bit more "finicky" for lack of a better term. So many knobs and dials to turn but which ones and how far. From my perspective I'm still not nearly as familiar with it as I was the Sabo but it'll all get there in time. I'm a big fan of the trial and error method, readin too much gives me a headache.

 

[ShrimpBrime]

Actually, was looking for a disfunctional AM3+ motherboard so that I can use the CPU socket to secure the CPU and then use a surface grinder to remove the nicket alloy and flatten the IHS.

Why not just Lap the IHS plate on a mirror and 400 grit wet sanding paper?

I don't think you need power tools. It would be neat to watch though!

 

[caddi daddi]

OH, I'm so with you shrimpbrime.

just get a whett stone from the sporting goods store, and the other thing you can do with one is sharpen knives, they are just so cool.

 

[mongoled]

but set it so the inlet and outlet ar across from each other rather than over each other

Dude, please can you clarify this !!



Thanks

 

[Johan45]

Dude, please can you clarify this !!



Thanks

Pretty sure he means to run the block with the inlet/oulet horizontal, intead of vertical. Side to side instead of up and down.

 

[mongoled]

Pretty sure he means to run the block with the inlet/oulet horizontal, intead of vertical. Side to side instead of up and down.

I tried that but flow was almost non existant.

I guess too much turbulence.

If he can confirm that it is indeed what he meant, i.e. to rotate the slot 90 degrees, then I will try again.

 

[Johan45]

Not the slot the whole block on the CPU.

 

[mongoled]

Not the slot the whole block on the CPU.

OK, I follow now,

something as obscure as that!

Wow, will give it a go



-- EDIT --

Forgot to say,

Ive managed to sort of stabalise the overclock at 5Ghz with the default slot.

I say sort of as core7 fails after 10 minutes but the other cores ran for almost 3 hours before I stopped the test.

This is using small FFT's.

Package temp was 64C when I stopped the test.

Ive now put my modified slot back into the waterblock and am running prime95 again.

At the moment its been running for around 10 minutes and core7 has not failed, water temps are 60C but core8 has failed instead!

With the original slot core8 would never fail, so the slot is at least altering the characteristics of which core fails.

I will rotate the block 90 degrees and report back

 

[Johan45]

I been thinking mongo that you said since switching boards your previous settings are stable correct. It could be that the CHV-z might be setting your ram a bit more aggressively and causing instability. Have you tried loosening the timings a bit to see if it stabilizes??

 

[mongoled]

I been thinking mongo that you said since switching boards your previous settings are stable correct. It could be that the CHV-z might be setting your ram a bit more aggressively and causing instability. Have you tried loosening the timings a bit to see if it stabilizes??

I dont think its down to memory timings while im testing small FFT's.

But to partially answer your question, on both boards I use the following methodology to set memory timings.

1/ Use the XMP Profile that is available for the sticks
2/ Then I go and tweak the settings that were/are required to be stable.

Are you suggesting that there could be some other memory timings that we do not have access to that the Z version is setting more aggressivly ?

Anyhow, as I will be modding the Raystorm mounting bracket I have put my Swiftech Apogee XT into the loop and have some interesting and strange results/characteristics.

Currently priming for abt 8 minutes, load temps are about the same as the Raystorrm, but with the Swiftech there are large spikes in the temp.

Currently the max is 69C, the spikes occur for a split second then drop.

Temps have now dropped to 57C from 60C a few moments ago.

Very strange .............

-- EDIT --

Well thats past the 10 minute mark and as per usual a core has failed and this time it is core6.

I dont know what to make of this, lol

 

[Johan45]

That's why I mentioned the memory and it's not so much timings you don't see as it is ones that would typically be left on auto. I can't get ASUS/ROG to load from work or I would link you. Find the uefi bios for the CHV-z at ASUS rog forum, it will give you a very good description of memory setting and what they're for and tips for setting them. There are a couple near the bottom of the Dram timings page read to read write to write and damn can't remember but they will get set too tight at time by the bios.

 

[mongoled]

That's why I mentioned the memory and it's not so much timings you don't see as it is ones that would typically be left on auto. I can't get ASUS/ROG to load from work or I would link you. Find the uefi bios for the CHV-z at ASUS rog forum, it will give you a very good description of memory setting and what they're for and tips for setting them. There are a couple near the bottom of the Dram timings page read to read write to write and damn can't remember but they will get set too tight at time by the bios.

Warning - visiting this web site may harm your computer!

Is being reported for the ROG site, lol

Will take a look .............

Cheers

 

[Johan45]

That's exactly what I'm getting??? It worked yesterday!

 

[mongoled]

OK, checked the timings, they are correct.

Will get the Raystorm mounting bracket sorted then continue testing.

Just a correction above, its a swiftech apogee xt not a storm, lol

 

[RGone]

http://www.google.com/safebrowsing/diagnostic?site=http://rog.asus.com/forum/&hl=en
Diagnostic page for rog.asus.com
RGone....

 

[caddi daddi]

try to rotate the block 90 degrees, play with the flow rate some, I had to slow mine back a little bt.

 

[caddi daddi]

rog.asus seems to be working now, here is a working link to the guide.

http://rog.asus.com/81732012/crossh...v-formula-amd-fx-8150-cpu-overclocking-guide/

 

[RGone]

I have hestitated to post...

...but now I think that I will. I am getting confused by some of the descriptions for temps. One time it is core/package temp we are referring to and then it is 60c water temp and I get lost in just what is how hot.

I am suspicious you are testing for inlet location in a manner not so likely to tell you what you want to know. Why?

Well the indications I get from seeing what is happening at 5.0Ghz is what I see when the Vcore is too low or the temp is too high. There is no stability. Some core fails. In most cpus that is usually a particular one first but not always and it swaps at times.

With the cpu so close or really not stable since it is dropping a core at 5.0Ghz, you need to drop back and get stable. Maybe 4.8/4.9Ghz or whichever. Then begin to change the entrance and exit modes into the block. You will know if you are better off if you can then lower the Vcore slightly and "still" be stable. Such a situation 'might' mean you are getting somewhere.

So far with the instability at hand, nothing you do is showing up as really better. You just seem to see a different core fail and that really may not point to anything. I can vary my Vcore enough when n0t stable to make a different core fail first.

Or there is another thing to consider and that is that changing the location of the inlet/outlet to the block makes no real difference. Or not enough temp difference that it can make a borderline not-stable processor suddenly stable.
RGone...

 

[mongoled]

...but now I think that I will. I am getting confused by some of the descriptions for temps. One time it is core/package temp we are referring to and then it is 60c water temp and I get lost in just what is how hot.

I am suspicious you are testing for inlet location in a manner not so likely to tell you what you want to know. Why?

Well the indications I get from seeing what is happening at 5.0Ghz is what I see when the Vcore is too low or the temp is too high. There is no stability. Some core fails. In most cpus that is usually a particular one first but not always and it swaps at times.

With the cpu so close or really not stable since it is dropping a core at 5.0Ghz, you need to drop back and get stable. Maybe 4.8/4.9Ghz or whichever. Then begin to change the entrance and exit modes into the block. You will know if you are better off if you can then lower the Vcore slightly and "still" be stable. Such a situation 'might' mean you are getting somewhere.

So far with the instability at hand, nothing you do is showing up as really better. You just seem to see a different core fail and that really may not point to anything. I can vary my Vcore enough when n0t stable to make a different core fail first.

Or there is another thing to consider and that is that changing the location of the inlet/outlet to the block makes no real difference. Or not enough temp difference that it can make a borderline not-stable processor suddenly stable.
RGone...

Hi RGone,

sorry for the confusion with the temps, I am always refering to package temps, it was a slip of the tongue when I said 'water temps' above.

As regards to stability, yes you are totally right, I am on the edge with 5Ghz Prime95, but Prime95 IS THE TOPS for me with regards to testing stabilty for CPU tests (not talking for graphics).

If I am boarderline stable with Prime95 for around 2 hours its means I am 'stable' within a high margin for any other CPU stress test program out there!

The reality is that I have simply switched boards and have lost that borderline stability I had.

Just trying to get it back.

All the feedback in this thread is proving very helpful and the aim is to be able to test the slot modification to see how it effects overclocking headroom

 

[mongoled]

Quick question!

The Raystorms slot, each side is not identical.

The slot on one side its inner is completely flat i.e. the metal sits flush.

Where as on the other side the slot has been filed down as so the slots inner is curved.

Which side should be against the copper plate as I dont remember ?



Ive also ordered another Crosshair Formula-Z as I am still uncertain that this motherboard is up to the standard of the non-Z I had before.

As all motherboards are not equal !