more 8370 results

Updating OP with current settings that are working.



old post for voltage compare

your package temp is right on the line, 62c is the tops i want to see and 55c for 24/7 use, I'm not sure about the later cpu's but the early fx got unstable at around 62c.
your nb is fine, you can slow it down, reduce the nb voltage and this will take a bit of heat out of the cpu.

if you're not bench marking, ignore ram speed, in game you will never see anything from it.

FX - OC - Loose - timings - 11-11-11- or 11-13-13 1.65v Getcha 1100mhz- NB chipset - 1.2v NB/CPU - 1.35v ++ - 2500mhz looks good -

4.75ghz or 4.80ghz = no difference - run the cooler temps.

Other than that, looks like a good OC. Low reference clock, nothing to worry bout there.

For the cooling, not bad at all.

Party on Wane!

Edit add: and I jinxed it. Checking the rosetta logs shows quite a few comp fails too, so looks like it's 4.75 then. Getting BIOS pics now.

It's up to about 5 hours now with rosetta without a computation failure, so tomorrow I'll reboot into bios and get the settings.

I did fire up some 3D to warm the video card up some and case heat - the CPU temp hit 66C which I expected to fail. But it didn't. Package temp didn't move much.

The card gets kinda warm - it's the "double die salvaged" Tahiti card, 7870XT which would also of been like a 7930. So it gets warm.

random plug - if you haven't seen Agenda Circling Forth - it's an "old" demoscene demo from 2010, it's pretty sweet. If you remember the point sprite test from 3dMark01 (the horse) imagine that on crack.
Agenda Circling Forth - Guru3D mirror

When launched it takes a while at the "Loading Disk 2010" screen because it's generating assets, so don't panic

And we have pictures.

case and cooling setup:
NOTE:
The BIOS/EFI pics are out of date - I had to increase cpu_NB voltage to 1.2625v for prolonged stability.
Also "CPU Bus Frequency is now 231 - not 232 like in the pics.

BIOS settings... or I guess that's technically UEFI settings these days.
Last one for fun. My workplace.

xrror, looking...

...looking pretty darn good man. THANKS for a bunch of pics since it really gives substance to the typing.

I think some of this has been said but will try and say it perfectly straight.

1. AMD tighter ram timings seems to help but faster ram beyond DDR3-2000 seems pretty pointless unless benching on sub zero cooling.

2. Fast CPU_NB helps if you are clocked above 4.8 or 4.9Ghz. So if you can lower CPU_NB so that you do not have to run more than 1.25V CPU_NB voltage...you can run a little cooler. Oh and fast CPU_NB is helpful mostly if you can run fast ram in conjunction with fast CPU speed. What I am saying straight out is if the ram speed is not really fast, you are not in CFx or SLi mode and do not have cpu running balls to the walls...save yourself a little heat by not worrying about how fast CPu_NB is beyond about 2400Mhz.

3. P95 Blend mode that kicks FX in arse and results in 65c or 66c on the Package/Core temp is not really bad on air since you really have to have a bad arse application to load the cpu that heavily when not in P95 Blend mode.

4. Now high CPU/Socket temps beyond 75c is not really good. Lots of air thru the case should be able to keep CPU/Socket temps below 75c. High socket temps can mean the VRM circuit is getting h0t and that is not really so good. I sort of look at the CPU/Socket temp as a way to know if I have pretty fair case air flow in and out since the VRMs can heat up under heavy load and I just don't like to see them get too too h0t. Fan directly on VRM maybe necessary if the CPU/Socket temp will not stay down.

Otherwise you seem to have it in hand. It is pretty hard to kill one of these rigs with the parts YOU have. So tweaking and testing and REtesting your tweak is where the REAL g00die and fun is found.


You might want to look this thread over. I went thru a pretty wide range of low to higher cpu speeds. You can get a glimpse at what and where the performance peaks and valleys seem to fall.

Good show man.

RGone...

I have a couple modern Lian Li cases. They are still mediocre as far as cable management and cooling options.

This was the smallest case that would still take a full size ATX board. It also happened to be on sale at newegg for $80 at the time so... SOLD =)

And this case ( PC-A05FN ) so far has been really nice to work inside of.

It also didn't hurt that I'm still really happy with the PC-62B (black) I bought in 2003.

EDIT ADD:
I guess I should clarify, the last system cases I've messed with recently was when I moved my total mess of a system (socket 1366 xeon with 2x120 AIO cooling) from that 2003 PC-62 - which pretty much had no concept of routing anything behind the motherboard (that era still had the cool removable mobo tray - of which no modern tower cooler could even fit through the back now) into of all things an IN WIN 904 Plus which... I don't even want to talk about that omfg.

seriously my workmates were laughing at me for a week as I had that on the bench, tearing what's left of my hair out trying to route cables through it so it didn't look like total $#(@

So basically yea that pretty much skews my cable management criteria in that "any" is still awesome to me

Found a scan of the CPU top before I put the chip in. I don't know if anything useful can be determined from the batch codes, but I'll throw it in here for reference.

when you crash at the LAN party, drastic measures come forth...

So weird thing. Got to take this to a LAN first time with this mobo/proc. Torchlight II. Computer didn't crash, only "Torchlight II stopped responding" and only when doing a LAN game.

This after playing many many hours of Torchlight II single player with no problems. Odd that. Makes it hard to test for - need more LANs apparently

so yea, ended up upping cpu_NB voltage to 1.2625 which brings a bit more heat.

and spring weather brings warmer temps...

and my air cooling was already hedging against the edge at 62C...

so this happened...



The cut wasn't the best, turns out my hole saw got "bent" more into an oval at some point. Which I didn't realize until after I had started cutting

I can get some HWMonitor shots if needed, but the short answer is this dropped temps:

cpu 56C
mobo 37C
package 53C

The one that surprised me the most was the package temp, it's lower now than when the front panel was off the case. I'm not sure how that works?

I also still don't know if this even fixes the problem. I lowered clocks down via mult (to 4.5, then 4.3) at the LAN as a temporary measure - but Torchlight still kept dying every 20-40min. And in single player I can't reproduce it (it always just works for hours) so this might be a false path.

So I'm guessing something on memory, memory controller, or mobo is getting unhappy.

But hey I have a genuine Lian-Li drink coaster now

more info as I get it

You can throw all the VCore you want at these things ..... well within reason but one thing still rules the roost. Care to guess .......

Yup cooling is KING ...... it's amazing what a little bit of cool air can do.

xrror said:

The one that surprised me the most was the package temp, it's lower now than when the front panel was off the case. I'm not sure how that works?

Click to expand...

Air flowing or moving thru a case, gives odd results when you remove the covers. Been a participant in such before. But key is the temp went down. Always a good thing.

RGone...

Wow these new FX 8xxx seem to really sip the volts!
Nice mod there, surprised temps dropped that much!

as the hardware wheel churns...

so this is a psuedo-update on this box. I ramble a lot, so I'll try and truncate. Also this beer is pretty good

Yeay, bullet points. 'ish...

- So a 16GB kit of gskill Trident X 2133 is on the way. It wasn't originally for this machine - yea that story is another thread...
but! more data points right? Not expecting any real difference from the overclocked 1600, but you don't know until you turn the knobs right?

- I have a screenshot of HWMonitor after leaving it open for 3 days. The temp flux is impressively wide. She gets hot when the building isn't A/C, but it *appears* to be running okay. I want that pic, as reference of what to expect of trying to run on air. But... I'll have to post that later today when I don't forget to copy the screencap to my email (d'oh).

- I need to grow a pair and just beat this thing with IBT and Prime95. Any suggestions of additional tests/flavors of prime95 welcome. Will torture on request. I should have done this first thing, instead of beating around the bush n00b style.

- I need to trawl Rosetta@Home logs for realz (tm) to see what errors I'm getting , but it's getting impressively close to my 1366 rigs - which being an AMD fanboi at heart, gives me a smile.

- I need to update the OP pictures, to reflect the actual settings I'm using now. Mostly it's voltage up on cpu_NB

- ask Silver_Pharaoh about coincidence of 231 HTT. Also on how long before his UD3 blows VRM

xrror said:

- ask Silver_Pharaoh about coincidence of 231 HTT. Also on how long before his UD3 blows VRM

Click to expand...

Haw haw

I haz fans over my VRM's
What's this about the 231 FSB I run?

I was just ribbin' you about the VRM's.

Silver_Pharaoh said:

What's this about the 231 FSB I run?

Click to expand...

I just saw in another thread where you were asking about different HTT speeds. By coincidence I ended up at 231FSB also.
When the new memory comes in I want to see if HTT speed (and NB speed) makes any significant difference in Rosetta@Home. Probably will take more NB volts than my air cooling can handle though

Sometimes it feels like GFX performance scales negatively to overclocks, but that might just be... or my ebay salvage video card ;p

The 3 day HWMonitor temps:

Ahh lol.

If you want that little but extra preformance, 240 FSB is my sweet spot on my FX 6300.

Going from 231 to 240 FSB gets me a 5 FPS boost in Heaven.
Running at my max FSB of 250, I get ~2 FPS more than 231.

About Intel Burn Test... what settings does everyone use?

Silver_Pharaoh said:

Going from 231 to 240 FSB gets me a 5 FPS boost in Heaven.
Running at my max FSB of 250, I get ~2 FPS more than 231.

Click to expand...

mmm, I like it... I see some tinkering coming soon. But also there is that pesky 240 giving greater performance than 250.
It's like the old days again, hunting for that synergy of timings, subtimings, ratios, etc for the parts you have.

Thar be performance in those hills, but gotta sharpen the axe first

But yes, with IBT what settings does everyone normally use? Maximum seems like it takes a looooooonng time. Which I suppose is the point.

I just did a run on "standard" "very high" and holy moly it lives up to it's name:

I also ran this on one of the gulftown rigs... somewhere over 85C it dawned on me that perhaps I should have blown the dust out first

EDIT: changed pic to "very high" IBT result. old standard result:

Standard is good yeah.
High is better.

Maximum is, well.... VOLCANO BURN IN
High is good enough if yer looking for stability

So the 2133 memory came in, and I had a little time to play with it, but nothing extensive.

First, I'm gonna have to break down and actually get a real video card for this thing - the 7870XT/7930 or whatever you wanna call it I'm pretty sure is tapped out so that caps most GFX benches.

The rough messing with IBT as a dirty/quick/already there benchmark (which probably isn't really a good thing to use as a bench) shows that memory speed and HTT makes a difference of squat.

Again, I only had a few hours to play today - and I need to get some *real* benchmarks on there past some rough stability testing.

Hopefully next week I can get some quality time in. Also I guess I can try low-rez Heaven.

More to come.

Also RGone has to be laughing - "hey look at this idiot try and reinvent the wheel" lol.

I suppose re-confirming all the stuff they've already done through my ineptitude still has value. Or at least some chuckles of "well duh" hehe