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

[TsunamiJuan]

So i am back with a new APU combo Once again going for the gold. I'll be posting progress in here as I go, as promised.

Sadly I am off to a rough start. After getting the initial setup done I didnt have any hard drives on hand so i messed around with some ram speeds a bit. Sadly I found that with the initial bios it had similar problems to the a75 fm1 boards had on release with ram stability. Timing tightness and speeds where limited for booting ability. Had sucessful boots up to 2133. Nothing in the 2400 range timing wise would boot, across 4 different sets of ram. Dominiator gt's 2400 and 2622's, and geil evo corsa's 2400, and RisjawX' 2400's.

After picking up some hard drives I installed w7 on it. Noticed that the drive controllers still exhibit poor performance outside of the OS. Yeilding Slow read and write times. I haven't tested this with one of my ssd's yet but It wouldnt surpise me one bit since the earlier hudson chipsets also had performance issues in these departments.

I ran out of time on the first night after installing everything. Since I had to head to work. So the next day when I came back I once again messed around with the board. After firing up the Easy tune software I found that the in OS core multiplier changes did not work. Also found the k8 stats will not recognize the cores for this apu. leaving that out of the equation for overclock. However the bclk was setable. Managed a mild setup overclock of 4.6ghz on default cooling with no load and no voltage changes. All voltages where locked on manual at the baseline accept for ram. Which was set to 1.65v per module requirements. (I was using my cas 10 ripjaws at this time 2x4gb).

I also found that while changing bclk frequency that cpu-z was freaking out on the ram tab, showing massive changes in timings to a degree that made no sense. (2-5-1 12 ht speed 12785) Which i attribute to be a mix of cpuz in-compatibility and bios instability. This was all on the release bios.

I had been having the bios bomb out quite a bit even on the release bios. So i decided to upgrade to the f3g beta bios. Which might I add I don't recommend. I had several sucessful boots and spend some time playing around in the bios again getting things tweaked and and re setup. I did manage to the the ram to boot at its rated speed of 2400mhz cas10. This require a marginal increase in northbridge voltage. and extremely small amount. Got back into the Os made an attempt at overclocking with the core multis again with no luck. Stuck to the bclk and managed to push the ram up to 2800mhz this time. (impressive) The failure point seemed to be inline with earlier apu experiences. It bombed out after the sync with the ram got unstable resulting in screen tearing and artifacting them a blue screen. Attempted to reboot and ran into corrupted bios image. after 6 reboots the image finally flashed back over from the secondary bios. (not the first time the bios has corrupted on this board, it seemed to be more tempermental about bios image than my old a75-ud4h was). Made some small changes here or there went back into the bios to try getting some stability out of in under it. Pulled of a clean boot then pushed for 4.8ghz. Got their and after two minutes it bombed out. Again resulting in a corrupted initial bios image again. This time there was no recovery. Flash back was incapable from the second bios. Possibily meaning that both chips in question where corrupted.

The hole time everything was under pretty much stock voltages. The error codes of 55, and b5 seemed to be purely memory releated. Despite that b5 indicates a usb host device hot plug problem.

I am gonna call gigabyte later today and see what they can tell me, aswell as any info they can give me about the lack of multi control. Possibly finess some hardware out of them to continue chasing some records with.

I also ordered another one of these boards since I probably wont see a replacement for the current one for longer than I would like to wait.

 

[dejo]

sounds like a long night of "fun". hate when all you get to do is fight bios issues

 

[Woomack]

There were similar BIOS issues on 990FX Gigabyte boards after going for high memory or NB clocks ( sometimes even on stock ). Really so what that they have 2 BIOS chips ... much better is 1 chip with recovery option and best would be like ROG boards have.

I wanted to get 2 core APU and cheaper GB board to check how far can it go but I don't know if its worth it and I think I will wait for some more results.
If it can make DDR3-3000 or more on air/water then I will probably decide to buy it

 

[Frakk]

Erm? I don't want to sound like a **** here, but in future maybe try ASUS Motherboards?

I don't know but thre seems to be a general feeling around that Gigabyte Motherboards are a bit...............Unpredictable?

 

[Woomack]

Erm? I don't want to sound like a **** here, but in future maybe try ASUS Motherboards?

I don't know but thre seems to be a general feeling around that Gigabyte Motherboards are a bit...............Unpredictable?

.. and if ASUS then you mean ROG series as all other are waste of time

 

[Frakk]

.. and if ASUS then you mean ROG series as all other are waste of time

Meh... ok.

 

[TsunamiJuan]

Erm? I don't want to sound like a **** here, but in future maybe try ASUS Motherboards?

I don't know but thre seems to be a general feeling around that Gigabyte Motherboards are a bit...............Unpredictable?

Theres a specific reason I didn't buy an asus fm2 board. The asus fm1 boards tended to push to much voltage to the socket fm1 chips alot of the time. In some cases this is good. But due to the unique way that the early apu's where designed with internal power gating and redisturibution. It usually meant that the amount of heat produced by the chip was higher.

I generally found that the gigabyte board that i had out of my three fm1 boards. Fared the best when overclocking. Now i admit i sold my f1a75-pro before i ever tried my 3870k on it with ln2.

As for asus right now, I am not to keen on their products. Since their switch to having ECS for manufacturing I have received more dead boards than I ever have from asus. I have also seen more shotty and unacceptable manufacturing quality out of them across the product ranges. So this leaves me looking to other manufacturers. Not to mention the possibility of any sponsorship from asus is quite low. I already have somewhat of a rep with gigabyte after their apu contests.

 

[TsunamiJuan]

Got another board today, same model as before, not the replacement board though. Haven't gotten a hold of gigabyte yet. Just been to busy so far this week. Have already spent the bulk of my week on the phone with work stuff.

The board got here to late in the day to really get anything done with it. Though I did fire it up. Ran into the same bios issues i ran into before with the f2 bios. Though the bios on this board seems to be recovering cleanly, and not failing out constantly.

Its kinda weird the a85x chipset seems to act alot like the intel p67 intels do on boot up. It does a series of prechecks, powers off then starts the boot sequence. Something which honestly gets annoying really fast when your pushing for the hardcore stuff. I'll probably set it up better tomorrow and burn it in somewhat. So that I know its stable before I start soldering on it so that I have some hard measuring points for my scope.

Might pickup a board by another company again. But I have a feeling I will be encountering the same issues I ran into with the a75 chipset. Since the bulk of the problems are likely to be shared across the platform.

Also I am thinking I will pickup one of Vince's venom pots for this. I wont have on in time for next weekends bench off. But it should still yield the colder temperatures that are ideal when it comes to AMD overclocking.

 

[ThorV8]

Erm? I don't want to sound like a **** here, but in future maybe try ASUS Motherboards?

I don't know but thre seems to be a general feeling around that Gigabyte Motherboards are a bit...............Unpredictable?

I am new here and don't want to start any fights but I have never had a problem with Gigabyte boards. From low to high end boards Gigabyte has never failed. I had to RMA three Sabertooths recently for a client of mine and a couple of Crosshair V boards. I think that ASrock and Gigabyte might make some of the best boards especially for OC'ing atleast this is my opinion.

 

[TsunamiJuan]

I am new here and don't want to start any fights but I have never had a problem with Gigabyte boards. From low to high end boards Gigabyte has never failed. I had to RMA three Sabertooths recently for a client of mine and a couple of Crosshair V boards. I think that ASrock and Gigabyte might make some of the best boards especially for OC'ing atleast this is my opinion.

when your pushing the limits with version 1 hardware you often run into problem. which is what i am running into. this is likely to happen with any new platform on release. the likely hood of most users having these problems are much lower. running 2400mhz+ ram is definitely outside the standard spec. Despite the fact that theres an option for it and i've proved the board can do it. Bioses are less than stable for it.

 

[Woomack]

I am new here and don't want to start any fights but I have never had a problem with Gigabyte boards. From low to high end boards Gigabyte has never failed. I had to RMA three Sabertooths recently for a client of mine and a couple of Crosshair V boards. I think that ASrock and Gigabyte might make some of the best boards especially for OC'ing atleast this is my opinion.

when your pushing the limits with version 1 hardware you often run into problem. which is what i am running into. this is likely to happen with any new platform on release. the likely hood of most users having these problems are much lower. running 2400mhz+ ram is definitely outside the standard spec. Despite the fact that theres an option for it and i've proved the board can do it. Bioses are less than stable for it.

I had no problems with Gigabyte boards till I got 990FXA-UD5. GB failed power design and they said it's ok ... when it wasn't. Similar issues were in some other series and for example some newer boards have problems to set higher memory clocks. Also their BIOSes were worse in last 2 years.

I never had problems with their cheaper series and actually H55 boards were cheap and great for oc, the same as some other series.
Their lower series are still much better than ASUS and I'm offering GB boards to most our customers.
I just can't see so much good opinions about their higher models especially that they are quite expensive.
For daily OC best seems ASRock what is my 1st option when I have no GB on stock in current price point.

For overclocking generally best are ROG series. Not without issues but results are best. You can also find some special series from ASRock or GB but not as full line of products with better support etc.

@TJ , I have hope that you find way to make higher memory clocks on this APU especially that your memory can easily hit 2800+ on air
As I said if I decide to try it then probably 2 core version. Hard to say is IMC will be worse.
FM1 series were quite good so maybe it's bios fault or again some power design. Early versions of boards usually have some issues and buggy bios.

 

[TsunamiJuan]

@Woomack, I am fairly sure that once the bugs in the f3g bios get worked out that higher clocks should easily be achievable. That was 2800mhz @ 10-11-11-32 i think, Hadn't played with sub timings and voltages much yet. What seemed to be the source of the problem is the lack of voltage on the NB. Which was often the requirement before on the llanos. So that you could push the higher memory speeds.

The good news is I shouldn't need to buy a two core chip. I will probably pick one up, I almost did the other day actually. But the next revision of gigabytes bios will include the ability to shut off cores from the bios. Unlike the llanos. This should help quite a bit when going for max clocks. Aswell as when driving the IMC for high memory speeds.

 

[Bobnova]

New alternate thread title:
TJ rips into a New APU - Kills it immediately


I am, of course, very curious to see how this works out.

 

[TsunamiJuan]

apu's are harder to kill than one would think. The boards seem to be the weak link. You where there last time I tried to kill one, and it still lives.

But yeah me killing hardware. Kinda a given. Can we say Cold fusion ice cream special?

 

[LiquidHard]

Good read for me, TJ. keep em coming

 

[TsunamiJuan]

Did some work last night. The new AMD over drive actually works for apu's somewhat. Not 100% but its better than nothing. Its kinda like mixing k8stats with et6. They kinda get along and they kinda don't. I am sure part of it is still bios revision problems.

The very interesting thing is that from AOD you can change the core multi, and gfx speeds. However the voltages dont seem to work at all, you can't change bclk. If you make the mistake of trying to mix changes in et6 after making changes in AOD it defaults the core multi down to 34-33x which is lower than stock. I am assuming it thinks the a10 is an a8 3870k.

Theres also a nice section for memory Timings under AOD. Haven't played with it yet. But its certainly nice. At the same time i confirmed that the f2 gigabyte bios in incapable of 2400mhz ram no matter how you try to pull it off. Really hope gigabyte gets a stable version of f3 out there soon.

 

[wagex]

in soviet mexico apu rips into you

does changing just the gfx speeds work on the non-k? my bios has an option but idk if it works, skeered to try lol.

 

[TsunamiJuan]

I am not sure if it does or not. I would assume no, since its suppose to be a party piece of the K models only. However you can still overclock the gfx speeds on both chips by just changing the bclk.

This is accurately reflected in GPUZ under stock gpu speed and actual gpu speed. For the new 10 chips it seems to be a 8x multi on the gpu. So bclk X 8 = gpu speed. This should be the general method that they follow for the APU's for gpu speed. I know some of the chips have a 600 and 400mhz igp speed. So it would make sense that they use a 6x and 4x multi in those cases.

I should probably note to that the a85x chipset allows you to change your NB frequency aswell. Which was something not allowed on the older hudson chipsets. Since it was a locked multi. The bios on this board will allow you to push it all the way up to 6000mhz. Where you can pull that off or not is anyones guess, but it makes for some nice headroom for overclocking. The standard NB frequency is dynamic aswell. It changes based on core and IGP loads. So if you have a high IGP load it will generally adjust itself to a much higher NB frequency. When all is working right this should yeild much higher performance at stock speeds. Much past those of the previous Platform.

 

[LiquidHard]

I should probably note to that the a85x chipset allows you to change your NB frequency aswell. Which was something now allowed on the older hudson chipsets. Since it was a locked multi.

1) Hi TJ, did you mean to say not allowed? Cos I cant find NB option in my bios, or maybe its jus this mobo.

2) CPUz does not show NB frequency on my apu fm1, the field is greyed out. Does it show on yours? Running cpuz 1.16 .3x64

 

[TsunamiJuan]

1) Hi TJ, did you mean to say not allowed? Cos I cant find NB option in my bios, or maybe its jus this mobo.

2) CPUz does not show NB frequency on my apu fm1, the field is greyed out. Does it show on yours? Running cpuz 1.16 .3x64

Yeah was suppose to be not, This is what happened when i get 5 phone calls in the process of writing a post. NB is not setable or Detectable on fm1. Its a hard set multiplier inside the chip.