Bitcoin Forum

Hi guys! I just installed this and wanted to check out if there are any performance improvements. Maybe we can all test this one out.
With my 5850 @ 960/300/1.174v | poclbm 6.27.2011 + aoclbf 1.74

Catalyst 11.7 = 389-393 Mhash/sec.
Catalyst 11.6 = 390 Mhash/sec.

Some guys reported improvements on OCL performance here http://bit.ly/iPSObE (http://bit.ly/iPSObE) (Post 41).

It's a new AMD APP runtime as well as driver.  Here is the download link also from the thread (thanks for users out there)  :)
http://bit.ly/kDIiWo (http://bit.ly/kDIiWo)

Enjoy!

What about phatk performance?

I just found out this.

http://developer.amd.com/tools/gDEBugger/Pages/default.aspx

The 11.7 Early has the SDK 2.5 in that package.

HD6870 from 284 to 292, HD6850 245 to 253.

CPU usage bug is still unfixed.

I'm very interested. Patiently waiting to see some results on the 5xxx series  ;D

5850 - 392MHash/s to 398MHash/s

What sdk/driver did u upgrade from.

Catalyst 11.5b

It seems like you don't need dummy plug anymore on Windows 7  :o

oh, good find.

Still getting the same hash rate but not having to use a dummy plug is actually a big plus.  :)

EDIT:  Seems you still need a dummy plug if you're overclocking.  Figures.  :(

not with CCC oc at least.

re the no dummy plus thing, does it work with remote desktop (RDP)?

didn't make a difference for my asus 5850

actually, it broke my machine when i restarted. now gui miner can't see my cards.

Hi! Maybe we can merge with this thread : http://forum.bitcoin.org/index.php?topic=24891.0. No dummy plugs? I'll test it on my dual-gpu rig. So far, no increase or loss in performance on the 5850. Maybe the 6800/6900 series have a boost?

my friend saw an increase ~2% with 2.1 sdk and 11.7 on his 5850/70. YMMV

poclbm 27.06.2011 comes with phatk instead of the old poclbm kernel.

Any news on the unneccesary CPU usage while mining?

So basically SDK 2.5 is looking to be as fast as 2.1.

No fix yet, at least for me.

Does 11.7 preview still hangs up on a 6950 if you launch a video while mining?

yes and no  i forgot to turn off my miner the other day and went to watch a movie and my pc dint lock up then i tried it 5 more times and it locked up every time so mabay we are moving in the right direction or it was just a fluke idk. but in the mean time turn of gpu gpu acceleration it fixes it

Does the 69xx mem downclock work properly yet? (I mean downclocking more than 100 MHz below core) Anyone that's tried it, I'd love to know - need to cool those 6990s!

11.7 Early + SDK 2.5 = ~755 MH/s compared to ~740 with 11.6 + SDK 2.4... (2x 6970s @ 880/170 [mem downclocked with MSI Afterburner after disabling EULA]  ...also disabled ULPS)

Mine seem to downclock well using Sapphire Trixx.

Going from 11.6 to 11.7 gave my 6990 a boost from ~360 to ~374. Nice.
No change on 5870s.

So, looks like everyone's reporting a small bump on the 6 series cards, but nothing on the 5 series cards?

Under 11.5 I've thought I've downclocked mem before, but I noticed that temps were increasing instead of decreasing. When I used some cli tools I found that the mem clock reset to max if I set to more than about 100MHz less than core.

Have you noticed a temperature reduction after downclocking mem? What did you downclock them to? This could be awesome - no more fans as loud as hairdryers for my SO to complain about!

I just killed my PC with a 6450 upgrading to 11.7. Went to safe mode and changed driver back to 11.6 and now can't even get into Windows. Safe mode hangs on storport.sys (how is this even related??) and Startup Repair from the OS drive hangs on a blank screen while Startup Repair from Windows 7 DVD hangs on Starting Windows with no logo.

And, starting Windows 7 setup from DVD also hangs...hardware failure?

what happens if you remove the card then boot up?

Same. Actually, I just got it to boot. Weirdly in my bios the CPU was undervolted by 0.3v. But the PC has been working fine for 2+ years (6450 is a recent addition). I had pulled out the 6450 and the eSATA pci-e card because of the hang on storport.sys but still could not get past the Windows boot up logo. The integrated graphics was also ATI/AMD so it would loud the same driver - I thought I was screwed. As a last resort I decided to reset the bios (this MB has always been finicky) and noticed the 0.3v undervolt! I wonder if this was usable before (can't remember) and just now at this point in time stopped working or somehow the bios got corrupted (seems so unlikely?).

I'm reinstalling 11.7 now to see if the CPU voltage setting was the culprit.
Anddd...the PC froze during setup.

Reinstalled 11.7 again without display driver and got BSOD..and here's the kicker, my CPU is once again undervolted 0.3v in bios. So now I have some software that can change my bios settings O.O

wtf? thats not posible, only with MB that has UEFI...

I don't have UEFI. I know it's possible to make real time changes to certain chipset flags but none that persist in the BIOS EEPROM. The undervolting may just be a quirk of the MB from all the hanging and BSODing...it's a weird MB. The system has otherwise been stable since I've put it together.

I have tried installing 1. complete 11.7 package, 2. 11.7 without display driver, 3. 11.7 with just display driver, app sdk, and VISION engine. All 3 has same problem - I'm suspecting the VISION engine setup now so going to try it without that.

Update:  Confirmed the instability is from the VISION Engine. Maybe it doesn't have the correct clock profile for my 6450 and thus runs it at unstable settings. Don't have an explanation for the BIOS undervolting except that the MB is quirky.

Nope, several mobos support tools provided by the OEM to runtime tweak settings, but only the expensive overclocker/prosumer ones.

I'm not sure if my board has those kind of tools but I wouldn't be surprised if it supported it. But the board has always been quirky - soft reboot (using soft switch) not always working, sometimes hang on cpu detect during post, etc. It's rock solid once it boots.

I wouldn't use the OP's dropboxed file when the third post has the real AMD link.

This release does not include a video driver, so it shouldn't be called 11.7 anything. The 110619a-121104E.zip includes the 2.5 SDK, but also includes ATI Hydravision and an ATI TV Wonder USB driver, so be sure to use the custom install option and uncheck those.

I got the exact same performance running this 2.5 SDK vs 2.4, on top of both 10.11 and 11.6 drivers.

The download from AMd's site does include a video driver. It is version 8.871.0.0 dated 6/19/2011 and is identified by VISION Engine as Catalyst 11.7.

I installed the sdk and display drivers only and I am not seeing a difference on my 5870. I've yet to test my 5850 or 5830 though.

I see what's I did there, the video driver is Win7 only, on other OSs it gives you no indication that it did not install a video driver (except that if you uninstall all ATI software and then install this you are still in VGA mode)

Win7 driver should be same as Vista. Are you running XP instead? Or possibly the driver is x64 only.

I first tested it on an XP miner, where the ATI installer did not offer any driver. The name of the driver directory in this driver package is W7_INF, so that should have given a clue that there is no XP driver. I don't know if it will install on Vista. You can always use the device manager and try to force the video card to use the driver in the W7_INF directory if the ATI installer doesn't install it for you.

I did test this driver release on Win7, and the CPU bug bit again - 100% CPU on one CPU core while running phoenix phatk. Going back to 11.6, and the CPU usage is back to 1%.

This driver lowered my mhash from 180 to 177 on my 5750 and from 117 to 115 on my 6570.
Sticking to 2.4 for now.

unless you're using only one GPU, I don't see how this could be true.

Tested it with two different cards, a 6950 and a 5830, and got no change (6950, Cat 11.6, SDK 2.4 -> Cat 11.7, SDK 2.5) to a decrease in speed of ~ 10% (5830, Cat 11.5, SDK 2.1 -> Cat 11.7 SDK 2.5).

Using Windows 7 on both machines, GUIMiner with Poclbm.

One CPU's usage is pegged at 100%? And yes, that Win7 machine only has one 5830.

I also have 11.7, and am dealing with the 100% CPU bug.

I hate this CPU bug

I switched to Ubuntu today because of the cpu bug - and now I'm happy (although I still have some other problems with Ubuntu - but miners are working without CPU load :) )  :)

It only effects Windows, and only some Windows users. Its similar to the SDK 2.2/2.3 bug and its caused by a broken busy loop that doesn't properly sleep.

Second, it would be very nice to find some sort of fix but I get the sense there just isn't enough demand.

Does anyone know if this 100% CPU bug is actually causing the CPU to truly run at 100% and drawing the extra power?  Even though I see the poclbm processes on my machines that have more than one card showing it's using 100% CPU, the systems are perfectly fine and responsive.  Usually on a system with a CPU 100% pegged it gets laggy, but this isn't the case with this 'bug'.

Now, if it is truly pulling all that extra power to drive the CPU at 100%, that can start adding up real quick...

Yep, cpu bug caused my sempron single core to max out and bsod from heat after 30 minutes.

Damn, that sucks.  So basically I'm wasting a few hundred watts of power running CPU's across my rigs for no reason.  Hope they fix this :P

And yes, I know I can go to Linux and avoid this, but I'm much more comfortable with Windows and the GPU over/underclocking capabilities.

Discovered a 2.5.684.212 SDK ;) But numbers seem to be the same.

Guys, I heard in this topic someone downcloked memory on 69XX. Still not sure how to downclock memory from 1375 to lower clocks on 6950@70 and on pure 6970
Tried to do this via MSIAF but it's still keeping constant 1375 clocks.
Same question for the overclock, trying to overclock my 6970 over than 950mhz via MSIAF and Power +20% immediately reset in CCC causing constant GPU mhz jumps from 900 to 980. How to hold Power +20% and OC more than 950?

P.S. No M/hash increase/change on my 2 6950@70 and single 6970 with 11.7 preview and sdk 2.4

915mhz - 393 m/hash
910mhz - 385 m/hash
950mhz - 410 m/hash

I had same problem as you on my 6870 but was able to downclock memory using clocktweak. However, there is no performance gain. If anything, it might be slightly slower.

Yeah, toss out CCC if you want to do any serious over/underclocking and go with something like MSI Afterburner and Trixx.

As for underclocking your memory, you won't see any performance boost.  What you will see is likely lower power draw and heat off the card.  That will sometimes free up a bit of headroom for you to give the core clock a bit more Mhz and boost your performance.

I've been running a 6950 (with the unlocked shaders Bios) at 880/625 without a problem and pulling around 390 MH/s.

Thank you for answers guys.
Okay, as I understand I can lower my memory clock to 625 only at default GPU speeds (880) and with uninstalled CCC? And if I'll overclock GPU more than 880 I'll receive constant jumping GPU speeds like 880/910/880/910/880 because of "Power +20%" increase lack, right? As I know I can control "Power" limiter only from CCC.