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

[LiquidHard]

man u guys r making me get fm2 itch..

 

[TeknoBug]

So what kind of OC numbers can you expect with just a Hyper 212+ heatsink and G.Skill 1866 ram and disabling the IGP? I've read 4.4GHz.

 

[watchman323]

TJ,
have you had a chance to play with the ASRock's board yet? Which BIOS has less bug for you to deal with, the Gigabyte or ASRock?

 

[TsunamiJuan]

So what kind of OC numbers can you expect with just a Hyper 212+ heatsink and G.Skill 1866 ram and disabling the IGP? I've read 4.4GHz.

I am not Familiar with that heatsink. Since I have never used one. You can generally do 4.2 ghz all 4 cores forced, with igp enabled on the stock heatsink. As long there is a good interface between the heatsink, Good air flow with a low average air temp. Atleast with the chip i currently have. Some chips are just better air chips than others. But so far it seems like the new a10's are more air friendly than the 3870k's where which pretty much hated air. Also more friendly than the 3850k's which Just didn't play so nicely half the time for a mix of reasons.

As for ram if your not pushing or running the IGP at all. There isn't as big of a reason to run super fast ram. The majority of people and applications will not notice the difference between 1600-2000+mhz ram. So it shouldn't be a problem. The only place it might hold you back is if your looking to overclock the bclk alot. Thats when ram thats atleast around 2133 is helpful. The multis only go so low. Eventally if your really really pushing things. You will be forced to run at higher memory speeds.

Now as for getting to the new boards. I am just now swapping lenses on my camera and setting things up to unbox the board. Had to clear off my memory cards aswell. I'll probably begin tearing into the performance and the differences between the boards tonight later In the evening PST. If I have the time. If not it will certainly be some time this weekend as thats the time That I have set aside to get the review completed (I Hope).

 

[TsunamiJuan]

So I tore into the asrock board yesterday. Starting the review spread on it. Sadly I can't update to the newest bios due to the bios image on asrocks site being corrupted. Generally I like the options in the bios though I am finding that its being a little restrictive on allowing me to do what i want right now. I'll be tearing into it more today. Out of the box the board worked with the ram at 2400mhz no problems at all, except when I go to reset it hickups on reboot with a memory error. Which I am fairly sure is fixed 1.40 (currently on 1.30). Sadly I have haven't been able to run at 2666mhz on the ram despite using the rated chips. Which is leading me to believe the either this IMC is not strong enough for 2600+mhz on air any longer, or the timings on this ram are still somewhat of a problem. Asrock I believe claims they ran at 2600mhz on gskill tridents. I might Pickup a set to verify this. I was originally gonna buy some but newegg was sold out when i went to order. Leaving me with the choices of avexir and corsair. I did however get awefully close to 2600mhz on air. Which is certainly better than what i was getting on the up4. Though I have since learned you have to use a different bank setup to achieve 2600mhz. the second bank not the first one. Which is certainly interesting. Though it makese sense now why gigabyte labeled bank 2 as bank one on the board now. (early revisions of the bios didn't like you doing this though, sighola)

Theres no sense in running 3d benchmarks right now, which is really what matters the most. Since I can't alott 2gb of ram to the video to put it inline with the other benches I have ran on the up4. Though its worth noting that at close to default configuration I pulled off a 5900vantage score. Which speaks highly as a starting point. I am sure it would probably be alot close to the other board's first run of 6200. If i had the same ram in, and prevented it from dynamic clock speed adjustments.

I'll probably end up pulling out the oscilloscope and doing some Ripple measurements aswell on the asrock board. Since It appears the some of the secondary power phases are single phases instead of duals, compared to the gigabyte up4. I am interested to see if this will make a difference when pushing the chips to its limits. Its highly possible that i might become a limiting factor. Though the a10's are much less power hungry than the llanos' ever where.

I also wanted to note that the ASRock windows based oc utility works correctly compared to ET6. You actually can change the core multipliers in windows, aswell as the bclk. Which speaks well for itself. Since both et6 and AOD have problems on a mix of these things.

 

[visiblegorilla]

Awesome! I was impressed with the windows based utility as well, I also love XFast LAN and XFast USB (and XFast RAM works just like RAMDisk). I'm running the 1.4 BIOS, must have downloaded it before the corruption happened. Also using banks 2 and 4, HSF blocked slot 1 anyway (oh well). Also using the Trident (2x 8GB) kit.

Great to hear you got 2600 out of your board and APU. I just realized that while I was able to run 2400 at 2.8+ vNB, I was using the XMP timings, which are faster than the RAM's spec (XMP is 10-12-12-31-43, RAM is spec'd for 10-13-13-32-41), so after I properly fix that, I should be a helluva lot more stable! Glad you like the ASRock, looking forward to your "ripple" measurements. Cool stuff. I'll hopefully have some Prime95-tested results at 2400 tomorrow with the new timings. Keep up the bleeding-edge work!

 

[TsunamiJuan]

Stabilitywise theres no problems with 2400mhz ram, I had lynx running earlier at 4.6ghz for a couple hours. Its far more abusive than prime 95 is, and will generally make memory problems show up in about 2-3 loops. Its also puts out more heat than p95 generally does.

I have to say that it once i figured out the weirdness of the bios wanting to keep down clocking the chip. The OC experience was generally easier on the ASRock board. 4.6ghz with no voltage changes required. The vrm's scaled nicely. I also noticed the the power consumption was higher on the ASRock board. However I haven't had to start fighting with the OCP/OVP yet on the asrock board either. Which was a source of frustration on the gigabyte up4.

It seems like the LLC on the asrock board is also much more effective. It makes sense though. One of the things I loved about the gigabyte board on the fm1 was that it wasn't as aggressive voltage wise as the asus a75-pro boards. Which tended to drive the chip to hard voltage wise. Which ment that it went outside the thermal envelope quicker. It was helpful when pushing on ln2. But it created headaches in other places. This is part of the reason that i attribute the gigabyte's fm1 offerings to be pretty much the top of the pyramid. Now that AMD has improved the power management massively on the new chips ( I am guessing that the bulk of the changes are internal since the major limitation was mostly internal before) you actually want a more aggressive vrm again. With nice strong LLC.

 

[visiblegorilla]

Wow, you obviously have a lot of experience with this stuff. The best I can determine is that the VRM on this board works very quickly and effectively. The XTreme Tuner software lets you watch the power phases jump around from between 1 and 8, granted it's software, so surely not "real-time," but cool anyways... LLC is set to 100% by default, I don't suppose that's something that needs adjustment?

So I'm running st 2400 again, fixed my timings (d'oh!), vNB at 1.3 again (1.28 unstable under load). How comfortable are you with vNB settings above this level? I feel like it'd be even more stable with more voltage, but I don't want to fry my new baby

Side note: The graphics run BEAUTIFULLY with the increased RAM speed. I'm really impressed. It's like I've got GDDR5 in there. Gotta love overclocking! Keep us posted!

 

[TsunamiJuan]

100% by default on that board is actually the reverse functionality on the ASRock board. If you read their settings int he bios it shows that 100% is default no llc at all. Then as you decrease the percentage it increases the llc amount. If the graph is correct that is. I didn't get into the power phases today. Was busy writing up other stuff.

Bios note, I got my hands on a beta bios that was just effectively pushed to all the sites over the weekend (became visible to me today, it either wasn't published till today or i missed it). So I am now running 1.42v Beta. However the performance issues I have been having persisted. This isn't a huge deal theres a work around for the problem. Its not a very enjoyable one but there is a work around. Spoke with ASRock about the problem and I am in the process of writing up a detail bug report for them. Since this bug exists on a number of bios revisions including the current beta. Their eagerness to fix this problem and acknowledge some of my suggested improvements thoroughly astound me. What I was dreading would become a multi hour hold session instead turned into about a 7 minute call. A real person picked up right away even none of this navigate the phone system.

Now that I've been able to confirm that this is actually a bug and not just a series of user induced stupidity, I can safely move on and start doing the benchmarks i need for a baseline and overclocking comparison.

Keep in mind that the asrock utility defaults to a higher dynamic voltage for nb than you would normaly guess. But its just that its a Dynamic voltage. It adjusts up to that voltage as needed. Unless your going through and forcing the voltages higher through the power state management with something like Tcik2 the voltages are generally gonna be dynamic. AOD does allow some of this control but not all of it. Generally the northbridge will autoadjust itself from the ranges of around 1.2v -- 1.28v, if your overclocking your gonna want that to go higher. It will somewhat scale higher than that automatically by calculating the approximate voltage for increased run states to some degree. Since the chips do have a few extra power state voltage setting put into them stock from the factory besides just the standard multipliers of 14x, 19x, 34x, 39x - 42x.

 

[visiblegorilla]

This is all coming together now. I too saw the graph in the BIOS and am sure you are correct, it is inverse. So "increasing past 100%" means reducing the value to 80% or lower. The voltage plot thickens...

As I posted elsewhere, I am convinced that ASRock wants to become the #1 motherboard manufacturer on the planet. That explains why ASUS wants badly to buy them out. When I bought this board I had no idea of the value they were including, especially for $99. I can't really say enough good things about ASRock right now.

Well I'm really trying to nail down this 2400 problem; I'm sure the G.Skill RAM is up to task (another company that wants to be the best... I've got a thing for underdogs, sorry Intel). So that leaves me squarely facing the Northbridge/Memory/GPU controller. Now, I may have a dud of a sample, but surely there is a voltage at which I should simply give up. I haven't seen anything above 1.3V in the wild, and right now I'm at 1.31 with ample air cooling... Can play Half-Life 2 all day, but no Fallout: New Vegas (yeah, I'm behind the times... It's cheaper). Where do I go from here?

 

[watchman323]

Visiblegorilla,
Asrock just posted a new BIOS. It included additional memory compatibility.

 

[visiblegorilla]

Thanks for the heads up. I've since figured out that I can run over 1.42 volts safely to the NB and get stable performance at DDR3-2400 in all my games. I'll grab that update, sounds cool!

TJ: Any news?

 

[TsunamiJuan]

Well i grabbed the new bios today, sadly they disabled spread spectrum *cry*. I messed around with it a bit while i was working on review something else. but didn't have a ton of time. Decided not to pickup any ln2, because its just a waste with how unhappy both bioses are. at 3-4$ a liter most of it would be wasted.

Haven't heard anything from asrock after i sent them the bug reports. I am guessing that it will show up in the next bios revision 1.60. Cause its a serious of rather limiting problems for those interested in overclocking.

 

[visiblegorilla]

I thought spread spectrum was kind of an outdated idea that hasn't really mattered since the 90's? I understand what it is, and surely every little advantage helps when you're pushing hardware to its limits. Do you usually get higher gains with it turned on?

 

[TsunamiJuan]

It depends alot on the platform and chip. I've had a few setups where having spread spectrum on stabilized the setup for extreme overclocking. Other setups I have had completely hated spread spectrum at high clock speeds. Its certainly something you want to have the ability to play with when it comes to overclocking. It also tends to affect the actual bclk value some.

For the trinity a10 you will notice that the bclk is almost all ways 0.1mhz lower than what its set or more as you start cranking it up. Once you start increaseing the bclk more in my experience so far it tends to push over to almost 0.8mhz.

Though interestingly enough I ran into the reverse of this when i was doing some memory overclocking yesterday with a discrete graphics card installed. The bclk was actually white a bit higher with spread spectrum on. Without going back and changing bioses I can't confirm if this is the case if it acts the same with it off since the options not available in 1.50v. But it was certainly an interesting find.

 

[visiblegorilla]

Visiblegorilla,
Asrock just posted a new BIOS. It included additional memory compatibility.

I have to say, this 1.50 update is super, super excellent because now I can run DDR-2400 at Auto voltages, allowing Cool 'n' Quiet to stay active while the BIOS raises vCore as high as 1.45 when needed, and vNB to 1.40 most of the time. Overall the board is requiring LESS peak voltage to remain stable now (vCore used to hop up to 1.47, and vNB needed 1.425V to play most games), so they worked some magic there too. Oh yeah, and the GPU core can run at about 1200 MHz now (1266 crashed pretty quickly under heavy GPU load).

Stock:
GPU: 800 MHz
RAM: 800 MHz (DDR3-1600)

Now (Stable)
GPU: 1169 MHz
RAM: 1200 MHz (DDR3-2400)

Soooo that's about a 50% overclock on air cooling. What ran at 45 FPS now runs at 60+ FPS. Cool stuff.

 

[TsunamiJuan]

Yeah i was messing with the bios today some aswell, I confirmed whats keeping me from hitting 2666mhz as being a lack of memory timing room, and figured out the other bclk stability problem. So I was benching some at 2590mhz on the ram, with the core at 4.7ghz ish on water.

Gonna try to straight it the rest of the way out so i cheeze the north bridge multi into working. Then go for large amount of benching. Generall if you up the voltages i really feel like the bulk of the APU out there will probably do 1250ish on the igp on air/water.

If the guys are doing 1600ish on ln2 with poor voltages and efficiency i dont see what you cant hit 1300ish on water or 1450 on phase change.

Plus i am not sure if you noticed this but the crash free bios recovery is now correctly working. 3 failed boots = quick rentery into the bios to make changes. At.east on air/water. Hopefully this is the same while cold. If it is ooh happy day

I am feeling like its worth it now to pickup some ln2. Once i get these baselines all out of the way tommorow.

 

[visiblegorilla]

Holy smokes, 2590 MHz on the RAM!? I bought the G.Skill 2400 cause I figured the APU wouldn't even DO that... That's really good news for this platform!

It seems like ASRock is working hard keeping up with these issues; I hope they're reading these posts, and your bug reports; whatever they're doing, it's working!

 

[TsunamiJuan]

I actually have a direct number for them now, so when i can finalize the evidence i need to work out whats limiting i am fairly sure they will be able to fix it with some bios patches. If its possible what i am asking them to do, it should unlock the pathway to push into the 2800 range stable.

Theirs definitely some odd quirks in this platform. Before Long I might bust out the soldering iron to see if i can work around some of the flaws.

 

[visiblegorilla]

LOL let me know how that works out for ya!