TJ rips into a New APU - 5800k on ga f2a85x-up4

[TsunamiJuan]

If you haven't seen this before, This is my soldering work bench, nuff said

Theres a few large tools missing like the a large circuit board heater, and the rest of the reballing tools.

When the iron comes out it gets serious

 

[Bobnova]

Results, as it turned out, were dubious

 

[TsunamiJuan]

Yeah what a waste of a zombie, damn card hated the cold, SO IT GOT THE HAMMER!!!!

 

[visiblegorilla]

I was wondering what the hammer was for... Now I get it. "Recycling."

And, hold on, is that a ln2 pot for a SLOT A?!?

 

[Bobnova]

It could be, now that you mention it.
It's built for GPUs though.

 

[TsunamiJuan]

Yeah the GPU pot could easily be mounted to slot 1 and slot a processors. It got more than enough mass for it aswell. Considering the wattage output of those chips is a fraction of what that pot is designed for.

I perfer to thing of the hammer as "persuasion" . "what do you mean you dont want to work, THE BEATINGS WILL CONTINUE TILL MORAL IMPROVES" *kachunk*

 

[TsunamiJuan]

While doing the OC benches for my review today, I thought my platniums died during it (not 100% if they are still working or not *sigh*) so I ordered a pair of TridentX's Cas 10, 2666's which are really the optimum chips for this setup. After talking with ASRock they confirmed that these are the same sticks they used to achieve 2600 in their lab.

This makes sense, the timings and sub timings on the cas 10 ram is still within the bios settings ranges for this board. Without rocketing out of what it can set. Would still love to have more settings but sometimes you have to work with what you have.

Also their is some interesting behavior out of the 1.50 bios now that i have spent more time with it. It somewhat remembers it previous CPU power states. So after setting it to say 4600 through the OC software after reboot it was coming back with the chip set to the correct power states of 42x (default multi setting that i had it set to in the bios). Which is quite interesting. Also somewhat found a work around for the NB, but its not a 100% fix.

Soon as the benchs are all done for this (i still have phasechange to do after i am done with water benches). I Will be ready for the real fun. Which will be later this week. I need to get some uber ln2 benches in before the end of the month. Not sure which board i will use but it will be one of them.

 

[TeknoBug]

I remember the Slot A days, my 1GHz Slot A got so damn hot and it was barely idling, my dual Pentium Pros got hot under load (~110-120C) but the Slot A got close to that without load.

 

[TsunamiJuan]

So After a long weekend of benching and review work I've Pretty much Killed the IMC on one of my chips. Or atleast its acting like its on it way out. What a bummer. Though Not to surprising considering the IMC on this chip is rated what appears to be the weakest by TCIK2 at -2. Which In a way speaks for the trinity cores quite well, if I was able to atleast boot at 2800mhz on the ram more than once at quite respectable timings. This kinda crippled things somewhat for the rest of the testing I had left to do. But no big Deal I ordered another two 5800k's today. Which I had overnighted. I thought about grabbing more but I've gotta leave some money for liquid nitrogen.

Seemingly the problem with the IMC don't show up as a pure instability so much as a high data corruption rate. You can get a consistantly set amount of time out of them at certain load levels before requireing a reboot. Before one has to reboot the system or it gets so bad it reboots itself. Which can be confused as an OCP or thermal overload protection reboot. It makes me wonder in a way if the NB on this chip was faulty to begin with since I've had data corruption problems on this chip from day 1 even at rated speeds (1866) Resulting if a 5D debug error on reboot. Requiring a hard reset. With more chips on hand I will be able to quickly work around some of these questions. Can never have to many parts .

As for review data I have a fairly complete list of data points now between the up4 and ex6 boards. Which I am in the process of compiling into a spread sheet. I Will probably add to this As I procure more boards. Even if they aren't review boards. It will be a nice comparision point as new bioses come out to show the performance improvements.

Will be picking up ln2 on wensday for sure. Leaving me two days To run these chips before the ECS fm2 contest @ hwbot. I'll probably toss in some entrys for the Gigabyte Fm2 contest at the same time. There is a good chance I will be live streaming the runs. So if your interested Check the Live benching thread on those days and see if something pops up from me.

 

[visiblegorilla]

What utility would you recommend to "see" data corruption in the IMC? I've a feeling that it's happening at 2400 even with the NB set at 1.45V. Heat is not an issue yet, and it's pretty darn stable, but it feels like there's a lot of errors flying around behind the scenes. Memtest86 doesn't produce any errors, I need a more powerful utility.

 

[TsunamiJuan]

Its not really a utility just deductive reasoning, after playing with the memory timings and speeds and seeing that I am getting memory related crashes. Though my perferred method of memory and stability testing is Lynx. Since its about as abusive as it comes. It will also generally get errors to show up quickly vs having to wait a few hours.

In reality its not so much a pure memory test as its a a full cpu test. Pushing both the logic and FPU of the chip to its maximum. It will also generally put off the max amount of heat for a normal processor. Not quite so much for APU's since you really need something thats stressing the IGP and CPU at the same time. Though you can allways run something like Kombustor and Lynx at the same time in these situations to find out if there are power problems due to OCP and TDP envelope control.

 

[visiblegorilla]

That's a great idea. I've noticed that Skyrim tends to put a lot more stress on the system than Half-Life or Fallout ever did. Let's face it, games stress the CPU and GPU at all times; maybe not 100%, but enough to let you know how your setup is doing. And I hear you, the seat-of-the-pants feeling you get when playing a game or running a benchmark under a higher clock speed that seems to instead have a negative effect... It's data errors. I can feel it in games (and measure it with Cinebench) when 1200 MHz GPU isn't as fast as 1100.

I've got Lynx, I guess that means I have to run it now. Damn, why don't I feel confident

 

[TsunamiJuan]

OOH man i feel like a proud papa right now. ASRock just released v1.60, And the mutiplier state bug is fixed and the HT multiplier works now aswell. Happy happy day. Now I have to conformal a second motherboard before tomorrow :\ O well twos better than none.

Though i do have a full 30L dewar of ln2 here though my other two chips still havent gotten here yet. But the magic is ready to happen somewhat.

 

[TsunamiJuan]

My dewar sprung a leak :\ so not to happy about that, but in about 12 or so liters of ln2 got alot of progress till i Either killed one chip, or ran into moisture problems. I was out of ln2 at that point anyways so there was reason to really dig into things more.

I broke down and called the local Gas place, They're bringing me a 160L dewar next week, i still have two more chips i havent even touched yet. Also I managed to get the chip to run 2666 stable so The IMC really just needed to go cold. Makes me wonder what It can do if I had a strong IMC. 3k+??

 

[visiblegorilla]

You've never told us what voltages you are using to achieve those speeds, so... I have to run 1.43+ Volts to the Northbridge to be stable at 2400 RAM, 1100 GPU... You?

 

[TsunamiJuan]

Nothing to extreme right now just 1.6ish vcore, 2.6v pll, 1.45ish nb. was at 1.8v earlier for the ln2 runs but i was on the UP4 not the ex6.

 

[visiblegorilla]

This air cooler seems more than capable of running 1.45V to the NB; there's virtually no heat increase from 1.38V. These grand overclocks are great for reference, because they define a ceiling for the rest of us modest overclockers, but what I'm driving at are real, sustainable voltages and clock speeds on this platform. 2400MHz is hard enough to pull off here. What it takes to do that is more than most people seem to be comfortable with. I don't care. This is an overclockers forum. Frankly if members tell me that 1.45V is too much, it's just enough for me. Let's not compromise here. What voltage do you need to use to sustain 2400MHz operation? Do you know of any other systems out there that run at these RAM speeds, and what NB voltage do they run?

 

[TsunamiJuan]

I wouldn't put to much into what someone says about APU voltages that has little experience with them. The simple fact is that APU's are power hogs. They will take a stupid amount of voltage and keep asking for more. The llanos acted the same way and so did the E series fusions on their initial roll out. I am sure at some point i will end up running the NB at around 1.8volts again. While 1.45 was kinda high for the NB on the llano, it was setup differently on that chip somewhat. Those chips also had serious temp problems. So a tiny bit of NB voltage would throw them quickly into the danger zone. That doesnt seem to be the case with the Trinitys.

As for sustained operation at 2400mhz, On the first chip i have despite the -2 on IMC quality it runs 2400mhz out of the box no voltage alterations needed, 1.15v NB voltage and its happy as a clam. I haven't Opened the other two Chips yet to find out what their IMC quality is. What you might consider doing is looking for another 5800k thats more suited towards air operation. Infact download Stilts utility and post the data from the first option for me cause i am interested in what the bin quality of your chip was to begin with.

With a air chip what your looking for is something with a strong imc, and low NB/IGP fused voltages. Low core voltage helps to, but i think those are mostly gonna be on high leakage chips from the few charts i have seen. For some reason alot of the high leakage chips seem to have weaker IMC's.

 

[visiblegorilla]

Yeah! Glad you're still awake. I can't believe you ran 2400MHz at 1.15 vNB. Inconceivable. As far as the binning of my chip, I really couldn't tell you, I'll have to wait until it's fried to give you the alphanumeric code (didn't write it down ). What's a "-2 on IMC quality?" How do you determine that? Sounds interesting.

I have no problem buying another A10, I'm actually looking forward to it, and will probably do so right before I push this one as far as it will go, I'm thinking 1.46 vNB might actually make the GPU perform at 1200MHz that actually beats the performance I'm getting at 1150MHz. Worth a shot with an extra APU at the ready!

PS I looked up "Stilts Utility" (and variants of that) and I got this: http://www.star.bris.ac.uk/~mbt/stilts/

Oh and by the way, are you using the XFast RAM at all? I like to run Waterfox Portable off it (http://www.softpedia.com/progDownload/Waterfox-Portable-Download-221223.html) as well as Minecraft combined with the Java x86/x64 installations, and it's stupid fast. Just wondering. XFast LAN seems to work well too. And XFast USB. I love ASRock.

PPS Thanks for the heads up on the BIOS. I swear to god I just checked yesterday.

 

[TsunamiJuan]

don't need the code off the batch number

use this
http://hwbot.org/forum/showthread.php?t=59243

option 1# and take a shot of the page that it displays. That will tell you the baseline info of the chip, which is what is generally the easiest way to bin these chips right now.

IMC Delta Strengh , is IMC quality. negative numbers are weaker, Positive numbers are stronger. I think it works on a -2 to 2 range.

As for getting more voltage to the IGP itself on the EX6 its easy just increase the IGP voltage multi in the bios. your probably gonna need to be in the 1.3v range for 1200mhz. (as opposed to what its running now despite the nb boost of probably around 1.275v @ the max igp speed powerstate)