Its sad I know but I've been clocking the crap out of the machine all night trying to beat a bulldozer Didn't manage it but did managed to find some new voltages, 1.8v through Phenom II Thermal cutout in a few minutes, it wasn't intentional, it seems my board ramps up the voltage by itself as I up the FSB Water cooled 1090T @4.33Ghz on AM2+ DDR2 board RAM @ DDR2 1050, dropping to 2 sticks gave me nothing that I thought it would perhaps I only have two good DDR2 sticks. 11194 http://browse.geekbench.ca/geekbench2/538922 Might have one more shot at it tomorrow, course numbers not as exciting as those headlining i7 but its all I've got, well I have an 8 core i7 laptop but its slow.
I hear ya sandys, it's horribly addictive just setting yourself little targets and making new ones each time lol. I beat a water cooled intel quad (QX9something) and now my target is 8k. I would probably romp to 10k if it wasn;t for that god damn awful AMD memory controller
Hmm i purchased geekbench and ran it, but I don't think it likes my 4x8gb modules, my memory score comes in around 4k which is incredibly low. i ran sisoft sandra to souble check and sandra has a huge difference in bandwidth etc. I've emailed primate labs to see what they say but i might go pickup 8 or 16 gb of ram to bench with this weekend
Out of curiosity, can you post a screengrab of your Sandra memory benchmark results? I'd like to know what effect such density has on performance.
If your memory isn't 100% stable, then you will loose allot of points...Try running the benchmark with stock ram settings.
Nice to see that AMD's Memory Controller still isn't quite up to par. On the flipside; I don't think it's entirely the Memory Controller. Could someone with an Asus Crosshair board try cranking up the Northbridge and HT link, while leaving everything else about the same? I suspect that might actually be where the bottleneck is occuring. (I'd do it myself, but my HT and NB max out at roughly 2.5ghz.) This is getting exceptionally annoying now. It's like this processor cannot handle 1600mhz memory. I always get errors, regardless of what tricks I pull, be they higher timings, more voltages, or lower clocks on the processor.
I'm noticing some really inconsistent memory performance between Bloomfield (920) and Gulftown (990X) With the same BCLK, mem and uncore frequencies, Bloomfield gets MUCH higher memory scores in the Geekbench test... but the memory bandwidth and performance is exactly the same in Sandra.
It looks like the voltages I need to achieve even a small overclock on my Bulldozer are ridiculous. I wound up having to use well over an extra 0.75V just to get 4.8 stable. Previously I could do that on my PII with only 0.50V. Memory does appear to be stable now that I've fixed the timings. Motherboard was trying a tRC of 34 instead of 41. Probably explains why it failed so much. Only needs 1.65V to keep it stable, too. Instead of the 1.8 I was pushing before. It also appears to be Prime95 stable. That's a nice change... This motherboard appears to grow dramatically unstable with a NB of over 2.2ghz unless I seriously crank up the voltage, too. That's a bugger for my memory performance.
What speed is your ram rated? Using a lower BCLK gives better OC and lower Temps. IIRC 5Ghz was 143 x35 with memory @ 2002mhz.
My ram is rated at 2000 C8 but will do 2100MHz+ C8 quite easily. BCLK shouldn't make any difference at all - the most important factor in memory performance is ucnore: higher is better, simply put. With memory at 2048MHz 8-9-8-24 and uncore at 4096MHz, I get only 6029 and 6136 with my 990X compared to my 920's score of 7763 and 8805 at exactly the same speeds I'll try 143x35 and see if it makes any difference... in theory it shouldn't, but theory is kinda out the window at the moment anyways lol. EDIT: no difference, exactly as I expected - you're getting much better memory performance in Geekbench compared to what I'm getting. I set 143MHz BCLK, 2002MHz mem (well remembered!) and 4004MHz uncore... and my score is still pretty much bang on 6K compared to your scores of 7180 and 7440... something is draining my memory performance with this chip and I haven't got a scooby what it is. I will get to the bottom of it however, damnit!!
What about employed AMDs tactics and disabling cores just to see how high clocks will go and go for a glory run Can you decouple your NB and HTT, for example my NB will run at 3.4Ghz (which reminds me I need to crank it back up) which has does wonders for things like memory score in this bench and gaming performance but the HTT becomes a problem at around 2.9/3Ghz, I have multipliers for this so can run it slower rather than tracking Cache speed. Also what about trying to run the bench with only 4 threads on each of the modules as Bulldozer isn't a proper 8 core, might yield you more performance as shown with some apps here. http://techreport.com/articles.x/21865
Decided to try and get over 6000 as I was so close anyway. Nudged it up to 3.7GHz... http://browse.geekbench.ca/geekbench2/539162 It's not rock-solid stable at that speed though, regardless of voltage and temperature it can't manage more than an hour of p95 if i'm lucky.
It's my NB. My HTT seems to be happy up to about 2.7Ghz from what I can test, but the NB starts flaking out at 2.2, And the HTT can't be higher than the NB. Disabling half of each module would yeild me 20% bonus in single-threaded tasks, or something silly like that, but Geekbench is multithreaded, I'd loose out in the long run. As much fun as the high clocks glory run would be; It'd be pointless for Geekbench scores. Oh, Pook; This memory can hit 1701mhz if you're a little generous with the voltages. Could you sit down and test what happens to the Memory performance as you increase NB and HTT speeds while leaving everything else at a set value? I'm really starting to think there's a bottleneck in there somewhere.
Tried it at stock and overclocked the figures are more a less the same, but sisoft shows a large difference. The score comes out lower than my 2600k and that's running 1600mhz on the ram and no overclock on the CPU. So guessing its an issue with geekbench and the large tam modules Sent from my HTC Sensation XE with Beats Audio Za715e using Tapatalk
There's something funny happening with Geekbench and memory on my system... can't explain it, but it's only Geekbench and no other software.
Looks like I won't beat my 11K score any time soon without a new motherboard, and I am not forking out another £130+ for a new AM3+ Motherboard that may need a Bios update to work anyway. (That I don't have the tools to apply.) Bugger.
OK I cracked the code and now I'm at the top of the Gulftown leaderboard, with much more to come - will try 5GHz+ when the cold air returns to Scotland My 4.7GHz submission, approaching 17K: Clicky
Well done LennyRhys! Although, you need to knock that vcore down to under 1.4v, I guess if you are running @ 205 BCLK, you might need more.
Yeah I could get it lower but didn't really tweak it much. The first time I ran Geekbench at 4.7GHz I got a BSOD with VTT at 1.3875v, but putting it to 1.4v sorted that out - the high BCLK and uncore needed the extra volts. The memory bandwidth killer has something to do with the CPU features, so I left them all enabled. C-States are disabled by default and enabling them further improves memory bandwidth, so I'm glad to have sorted it out!
http://tinypic.com/r/bip7hk/5 heres my best run so far, just cannot hit 12k, my memory is letting me down i think, when i go above 210 fsb i cannot get a good memory divider, it either wants to be at 1200mhz or 1700mhz, the most i can get on my memory is 1686mhz.
