Formula350

Yea, honestly though, that doesn't really surprise me given that I think CPUTweaker was probably the last fan made/third party utility that did memory timings on AMD chips other than Overdrive, and as far as I know it had ceased to support AMD after Thuban (not even my Thuban-based Stars core A8-3850 is supported by it, which was a real bummer). For that matter, short of being either really good with figuring our registers along with having the tools to manipulate them, or the industry connections to have the awesome insider-programs, then seems like you're basically SOL these days (especially when it comes to AMD ). I remember when the OC record was set on Bulldozer and I was looking at that screenshot wondering "Man, I've not heard of PSCheck, it looks awesome!", then much to my surprise managed to actually find it on AMD's site (along with a couple other programs that I nabbed, but long since outdatted lol), I was one happy camper! Sadly, I couldn't contain that excitement and shared it with everyone... thus, that area of AMD's site once again getting closed to the public Oh well, right? I did see a thread on, I think Overclock.Net, where someone had said they had updated MSRTweaker for Carrizo, but for the life of me I can't find that thread again :\ If anyone happens to know what I'm referring to and can point me to it, I'd love to try to message the guy. Or *cough* better yet... if someone with coding savvy is able to update it themselves for us all, hell that'd be even better!! :ws:

Desperate times call for desperate measures... in a manner of speaking While I'm indeed in the neverending pursuit of more performance, in this instance I'm looking to take the "Polaris Approach" with less=better, as what I'm sporting is the mobile FX-9800P The most I've been able to come up with for anything tweaking related is an unreleased program made (modified?) by The Stilt, which he (understandably) is unwilling to release publicly. I've tried the most recent AMDMSRTweaker I was able to dig up which is v1.1 (and another one similar version, but is curiously much smaller file size [no, not 32b vs 64b heh]), but from what I've been able to tell the changes applied do not 'physically' result in any changes. The readings are reflected MSRTweaker, but nothing seems to actually take hold, at least as far as CPUz (and similar read-out software) or benchmarks detail. I know that the Kaveri and later models have issues with support, namely voltage related which is present of course, but even when dialing back that number which was provided there doesn't seem to be any changes. Thermals under stress testing seem to be the exact same as before making any voltage adjustments. Also, the PStates don't appear to be holding the multipliers either, as I've tried to make the Boost multis a lot lower in order to keep a more consistent clock, but everything appears to indicate that clocks are unchanged. Reasoning? Well beyond the obvious of increasing battery life, and lower heat output, thus hopefully increasing graphics clocks in games, it's also because the way that AMD chips are setup and in turn handled by Windows power policies... sucks. :\ For example the vast majority of the percentage values do not function at all, due to a large range covering most PStates, such as with the lowest PState covering pretty much everything from 0% to 55% Or the fact that 100% = boosting, regardless, and 99% = not full-speed but the next PState down, which is like 2.2GHz instead of 2.7GHz. Beyond that, I sure do miss the days of in-windows RAM Timing tweaks What I wouldn't give for something like A64Info again! At any rate, I doubt anything will come from this, but I honestly couldn't think of anywhere else TO look. Thanks :celebration:

No prob. Seems that the other A75 motherboard BIOSes are waiting on my testing of this one's, so I don't have anything to upload as I originally though. So I'll fire word over and let him know of the issues I've came across, but that the CPU-NB divisors work up to 900MHz at least.

Note: As far as I'm aware, this is only for the A75-UD4H. I titled this thread with only "A75" as I am hoping to get the same Beta for the -UD2H (no guarantees). Please read the included Disclaimer text. It includes what has changed, and what I've so far discovered as bugs.

Hmm... <_> Well, alright, but I'm including a disclaimer!! lol

Noticed a hugeeee difference in memory performance between the ASRock A55 Pro3 and it's replacement, Gigabyte A75-UD4H. The ASRock was getting anywhere from 2000-3000MB/s better Write speeds, around 10,000MB/s to 11,000MB/s. The UD4H was only managing 6,000MB/s to 8,500MB/s (I did break 9,000MB/s in an unstable config though). Initially I thought it was a problem with the CPU-NB (Mem Controller) in the BIOS, which I discovered was not running at it's full 900MHz at any DDR3 speed, which it should auto configure it 900MHz when running at DDR3-1866. Instead it was stuck on 720MHz, which is what DDR3-1600 auto configs to, and that's a healthy speed difference! Informed Gigabyte and they got a beta BIOS out to me, which it now provides manual NClk (CPU-NB) divisor control, but Auto is still bugged and not running at 900MHz. Regardless, you can select a divisor of 4.00 and that is 900MHz, so everything is peachy there! Hopefully they'll release that BIOS soon (my board may have died though so I can't continue testing it ) However, that unfortunately didn't resolve the problem with the Write performance. While tinkering with the F7 release, I had decided to "Enabled" the "C6 Support" option (Default is "Disabled") and what do you know... FIXED WRITE SPEEDS! Overclocks don't seem to be effected by Enabling C6 either, so that's good. To enable C6 Support, just go into the Advanced Chipset BIOS menu and it's the second option in the list Simply cursor down to it, and press the + key on the keypad or hit Enter and select "Enabled". Back out and goto "Save Changes and Exit" (or Press F10 from any menu) and that's it, right as rain (To those wondering about the Beta BIOS. There are a couple new bugs in it that could cause problems for some. The biggest, which will result in your system not POSTing and requiring a CMOS clear, is the selection of any NClk Divisor lower than 4.00 (900Mhz) -- IOW selecting even 3.75 (960MHz) will require you to clear it. Also, the Memory Timing Menu's Proc ODT's "Auto" option does not work, not sure if it's broken any further, but it jumps to 60Ohms and doesn't display as "Auto" if you select it. So as a result, and not knowing Gigabyte's policy, I don't feel right with uploading it anywhere, I'm sorry. )

Been testing a Beta BIOS (F8b) for Gigabyte to fix a huge issue I discovered: CPU-NB (NClk) not running at full speed (runs at 720MHz, even when set to DDR3-1866, instead of 900MHz like it should) and effects what I assume to be the whole A75 line. For sure the UD4H and UD2H. Anyways... There is now the option to select the NClk divisors, up to quite an extreme level, though unfortunately anything lower than 4.00 (900MHz) results in the system not posting, and not recovering via Backup BIOS. But still, it's so much nicer! I'd attach it, but I'm not sure how GB would feel, and also I think the rather early (slightly broken) NClk adjustments might cause some undue tech support tickets Oh forgot, the memory Proc ODT is either glitched or broken too, with AUTO completely not working.

Tried it with the latest version? CPU-Z 1.59.4 just got posted on Stasio's list of tools today. I haven't had my desktop connected to the network lately so I can't tell you if it's approving results.

ASRock A55 Pro3, A55iCafe, A55M-HVS. ASUS F1A75-V, F1A55, F1A55-M LE, F1A55-M LX, F1A55-M LX PLUS, F1A55-V, F1A55-V PLUS MSI A75MA-P35, A55-G45, A55-G35, A55M-P35 Just if you want a more-complete list of boards that are out there, that's all

Last post of images for now, unless someone has any requests. (long pics, sorry, had to combine 2 into one) MaxxPi2 Pro (internal beta) "Memory" bandwidth scores, highlighting again the vast difference in memory performance between the two boards. This time I went and configured the Gigabyte to the same settings as the ASRock. The only thing different (from what I can change) is that the Gigabyte has IMC Ohm, Drive Strengths and Drive Timing options, which I leave at Auto. --------- MaxxPi2 Pro "Overall" memory score, same as above. --------- Best Sandra I achieved with the ASRock, and at the same (stable) speeds as the above (3.58GHz, DDR3-2300). I've yet to reinstall it to run on the 3870K (went all screwy, won't run) while overclocked, but the default config only net like 16GB/s on the Gigabyte --------- Sandra, again, but just outlining the change in performance on the ASRock board between the default NB speed of 900MHz when you set DDR3-1866(which I listed as it's rated 1800MHz) and the 742MHz (1500MHz) that the system selects at 1333. Also is the 2300Mhz which results in 1112MHz (2224MHz). This was before having done the subtiming tweaks in the above results. --------- ASRock's BIOS showing the option for CPU-NB speed options. It's a drop down listing a whole bunch, but no screenshot of that, sorry.

Random Cache-Mem results of the 3870K on the Gigabyte. The second one is the mate to the previous post's 1866 6-8-6 results. -------- 3.9ghz cachemem.png Results on the Giga with 3870K@3.9GHz, which I dropped the DDR speed down to in turn lower the CPU-NB (which sadly I have no choice over) to rule them out. Interestingly, despite those drops the Write performance isn't as bad as expected. Couple more left, YAY 5 pics-per-post! :banana:

A55 Pro3 - AIDA64-3850@3.58+NB@1107+DDR3-2300.png This is the best, stable results with that board. It's only a 3+1+1 phase config by the looks of it, which after having the GB board my suspicions that it was limiting the overclocks is confirmed. I can bench the 3850 at 3.9GHz and DDR3-2445, nothing that ASRock A55 board could ever dream of doing. ---- AIDA64-3850@3.8+NB@943+DDR3-2445.png Memory score for the Gigabyte, the best I've been able to manage so far. Quite good, but it's Write performance sucks, which again, is inline with what I experienced on the ASRock board when I disabled the ASMedia SATA-III controller . Not sure why that would occur, but it was consistent, and even then the Write was about 700+MB/s faster than this (fastest) Gigabyte result :\ ---- AIDA64-3870K@3.0+NB@720+DDR3-1866.png Defaults on the 3870K in the Gigabyte using the timings I've been using as "default". This is using the foolishly low CPU-NB speed of 720MHz, which doesn't help performance as you can see... ---- AIDA64-3870K@3.0+NB@720+DDR3-1866 6-8-6-34 CR2.png Defaults on the 3870K in the Gigabyte, lower timings (forgot to adjust the Command Rate and tRAS lower) and crappier than expected performance ---- AIDA64-3850@2.9+NB@900-DDR3-1866.png 3850 on the ASRock, as you can see, Write performing much better at defaults with NB@900 than the Gigabyte does at any NB and DDR3 speed combination ---- On all the AIDA64 compilation results you can see that the board doesn't much effect any of the benchmarks which have to deal with just the CPU, which some are only impacted by CPU speed, and so the higher NB or DDR3 speed doesn't make a difference.

It is Hidden, but it still doesn't work after restoring it That's what I mentioned having already done by restoring it in F7 (which PM me if anyone wants that, not that it seems to function). I'll try rolling back to F3 and see if it's able to control the CPU-NB. Fingers crossed, because I don't like having one of the best boards and not having such a key option/setting to change... Going to gather screenshots, I'll post them in a separate reply.

ALRIGHT! We're back in action thanks to a thread Stasio linked me to Holding the power when turning it on so that it'll then turn right back off in 4 seconds, then powering it back on, will trigger the second BIOS to engage (After it shut off I left the power button held in for a few more seconds just to make sure) There is not a multiplier, but it's a speed on that board. It is only selectable up to a certain MHz at a specifc memory speed. So at 1333 for example it'd limit to 720Mhz (which the Gigabyte runs at under DDR3-1866!), and at 1866 it runs at ~900MHz I've got screenshots (which are in my review I linked to earlier), but it's almost 2:30AM, so I need to get to sleep. I'll post up a bunch of stuff tomorrow.

I let it sit powered on for about an hour and I could hear it (in the background) in a reboot loop ever 5 seconds or so, which coincided with a flicker of the HDD LED. I tried everything to get the system working, but the USB slots don't get any power to do a Recovery Flash via thumb drive, the Optical drive isn't being accessed for a Recovery Flash on a Bootable CD and the shorting of BIOS pins I tried but I have no idea which pin/s to do since it's far different from the board the original trick worked on :\ Yea it was a borked flash, which I know it should auto recover, as my 890GX has done it for me but it isn't doing it on this board Didn't help mine any. That is just for helping memory overclocks, at least from everything I've been reading. Eitherway, the Pro3 was always enabled, and it was the Disabling of the add-in SATA-III chip (ASMedia) that is what caused the system to exhibit crappy memory Write performance. I went through every setting I could, even going so far as to disabling SBGPP, but even that didn't help :\ Anyways, I did manage to get some benchmarks ran to illustrate the difference in performance, so I'll plug that system's HDD into a USB adapter to grab them. If anyone has any thoughts on which pin to short to trick a recovery, and perhaps WHEN to short it, here is the pinout:

Awesome... The backup BIOS does _nothing_ on this board! :( On my 890GA-UD3H board if I had a bad flash, the board would boot up and give me some message about the Checksum being bad, then recover from the backup. This board... Not a thing, doesn't even POST *cries* Went from a dead board and not being able to review the 3870K, to now a bricked board and will have to wait a few weeks before getting it back

The ASRock board unfortunately died on me (power phase area I believe) so I can't get any more data or screenshots from it. All I have are that which made it into my review and the various tidbits that didn't. So if you were thinking of something specific let me know and I'll see if I happen to have that. As for the UD4H, I'm open to supplying whatever I can! I'd love to get it on par with the Pro3's write performance, as that really puts a damper on my mood lol We bought it to review the 3870K, and I was hoping it would be a kick ass board. Which so far in most regards it is, but the BIOS is not what I expected (though I knew it wasn't UEFI), and just strange it is being outperformed by a budget board more than half it's price Does great with overclocking so far. I'm still learning the quirks of the APUs, but it'll make it to the Windows boot loader at 4GHz, also at DDR3-2500, but that's it. Dunno if I can get the DDR3 stable at that since there are issues with something (board or modules) when setting timings higher than Cas 11. Won't POST at 12 or 13 (thus, won't post on Auto timings since it's 13-13-13) *shrug* I'll also look into messing with storage controller settings to see if it magically effects performance like it did with the Pro3. Suppose I should go fiddle with things now... I'll report my findings. Let me know what sort of screenshots you're interested in.

haha Is the Honda City considered a van? Mine, then, is technically a 1993 Firebird Formula 5.7L (350) :nana::banana: To stay on topic though... lol Got F6 downed and will give this a go a bit later; stuffs to do! I assume there is no F8 Beta/Alpha in the wild yet?

What do you mean by "drive"? Hard Drive, optical, memory drive strength That's strange and doesn't make much sense to me :\ At least by traditional thinking it doesn't. I see it as the old CPU-NB that it just can only be set so high before you lose stability or the system POSTing, so on locked chips you need to be able to lower that in order to keep it in a stable speed range. Yet on the flip side, if these chips have a CPU-NB that simply doesn't have a limit at which OCing the FSB/HTT will reach, then I can understand removing it; however, there is still the issue of Gigabyte locking it at only 720MHz, even while running DDR3-1866 where it should be 900MHz... There shouldn't be any reason to not be able to change it on these (mostly) unlocked A8-3870K chips though, so removal of that option is a bit odd. You can change it on my 3850 though, as I was able to on the ASRock A55 Pro3, though that didn't offer multipliers. It's method was rather user friendly by automatically setting the speed to the max when the memory speed was set as so: (these speeds are not verbatim as I always ran at 1866) At DDR3-1333 it only allowed a CPU-NB of 720MHz or lower At DDR3-1600 it only allowed a CPU-NB of 840MHz or lower At DDR3-1866 it only allowed a CPU-NB of 900MHz or lower So the ASRock A55 Pro3 at default FSB while set to DDR3-1866, the CPU-NB speed was 1800MHz. I think I figured that either that board's FSB limit, or my memory's, was around 124MHz. That ended up giving me a CPU-NB of 1116MHz, or 2232MHz effective, and a DDR3 speed of 1153MHz (2306MHz effective). However, the UD4H (so far with my modded F7) while at 131MHz FSB only was running the CPU-NB at 943MHz, or 1886MHz effective, with the DDR3 running at 2445MHz! That CPU-NB is about what it should be at the default DDR3-1866 :\ I'll grab F6, first flash the unmodified F7 and test, then flash to F6 to see what's what.

I was going to post this in the BIOS thread, but this seems just as fitting (it not more) a place: Has Giga not included anything for changing of the CPU-NB multiplier, or did my modding of the BIOS to always make it available break it's operation? I unhid it's option so it is always there because the speed is not correctly set on these Giga boards it seems (UD4H specifically). I had the ASRock A55 Pro3 previously but it died, yet it had the CPU-NB speed available to set even on my 3850, and at DDR3-1866 it would auto set to the highest available (900-some, or 9x IOW). Even at 1866 on the UD4H, it is only running at 720MHz (7.2x multi), which AMD has said they set up the APUs to run almost 1:1 with the DDR speed. So 1866, or 933 actual, should be running at 9xxMHz CPU-NB, even on the 3850s. That's what I modded the BIOS for, to include that, but it's function doesn't impact the speed any I tried lower than 9x down to 5x, but it was always at 7.2x! Even with the 3870K I can't seem to adjust it, but I've not tried the unmodded F7 to confirm that. Also the damn Write performance on this UD4H is HORRIBLE. The most I've been able to get, which is at DDR3-2300 BTW, is a pathetic 9200MB/s... With that ASRock A55 Pro3 it's an amazing 12000MB/s!!! That's a huge difference! That's in AIDA64, and in Sandra the Pro3 was 22GB/s, with only 18.5GB/s on the UD4H One thing I did notice on the Pro3 though, is if I disabled the additional ASMedia onboard SATA3 controller, my Write performance took a huge dive, to exactly where the UD4H is at Why on earth that is, I have no idea, but it's rather strange. All across the board that Pro3's memory performance is more than the UD4H and that doesn't make sense IMO heh UD4H obviously OCs tons better though! Also not exactly sure about the temperatures. I'm using a VERY capable cooler and yet the board (through Windows) is reporting reaching up and over 70C, depending on voltages... On the Pro3 and a much smaller heatsink I don't think it ever reported breaking 55C (at 3.5GHz, 1.42V), but it easily breaks 60C on the UD4H. That's even with a fan blowing directly on the back of the motherboard :\

That's fine haha It'll only be worse when run at 1280x1024 (external monitor) instead of 1280x800 (laptop's screen), not that it's possible since all the GPU specific scores were already "0.00"

Check.... Mate http://hwbot.org/community/submission/1072792_formula350_3dmark_vantage___performance_gma_4500_q45_q43_0_marks

Since I can't edit my submission for whatever reason... This is actually an Athlon II X2 250u (ultra-low voltage) and not a standard 250. They are the same in basically every way I can gather, except I'm locked w/ a CPU multi of 8x : 1.6GHz is stock speed, so I had to crank the HTT up to 350 to get into any competitive performance! The best trick I managed to come across was setting the UMA SidePort and running 512MB for UMA. Don't know if I should've set the DDR2-SP speed to be faster than my DDR3 speed, but I didn't. In hindsight I should have, but oh well! I was only running Win7 32bit so I set the UMA to go above 4GB and that way the unused 512mb of my memory was used instead of eating into window's memory, not that 3DM2K1 uses much. Had my modded 890GPA BIOS worked, I would've set the UMA to 1GB, but it kept kicking a Checksum error So if anyone can do it with 1GB, I say give it a try! Not that I use it, but I turned off NB Azalia (IGP's audio) just in case. Can't say for sure it did anything or if it was just the extra voltage, but I remember initially only being able to do 995MHz core, where as I pulled that bench off @ 1075Mhz (I did manage 1100 but it started to become unstable).

Side note: Wish I was at home to enter in new scores since I have a 555BE waiting for me that does 4.2GHz on air Oh well though haha I was happy to have 3rd place for a day on a 250u (ultra-low voltage, 1.6GHz stock) @ 2.8GHz =}

And no offense either, but this is a friendly competition with no prizes, from my understanding It's just a "have fun" and "ePeen" booster