GPU comparison (Neoscrypt)
-
Visjay is that the prebuilt one for Linux?
Edit: Stupid question just spotted the c:/ windows at the top on the window doh!
-
In my post back a page I forgot to say about the driver used> this is with AMD driver 14.9 using a 14.9 .bin, as the 13.12 .bin gave the same results/hash rate with no change when in use with 14.9 driver only by deleting an let it make a new .bin for the 1.4.9 driver did I see any improvement in hash rate as shown in the picture above.
My old hash rate was 37kh/s per card now 60kh/s per card with 3.7.7b and 14.9 driver/14.9.bin in OS W7 64bit.
For those looking for a pre-compiled 3.7.7b for Windows it can be found in post #224 here
there are also other pre-compiled 3.7.7b in that thread some on that same page I believe…
2xR9 270 non X
-
In my post back a page I forgot to say about the driver used> this is with AMD driver 14.9 using a 14.9 .bin, as the 13.12 .bin gave the same results/hash rate with no change when in use with 14.9 driver only by deleting an let it make a new .bin for the 1.4.9 driver did I see any improvement in hash rate as shown in the picture above.
My old hash rate was 37kh/s per card now 60kh/s per card with 3.7.7b and 14.9 driver/14.9.bin in OS W7 64bit.
For those looking for a pre-compiled 3.7.7b for Windows it can be found in post #224 here
there are also other pre-compiled 3.7.7b in that thread some on that same page I believe…
2xR9 270 non X
I suspect that with that on a pool, you’d get 100% hw errors… When I dumped the bins from the 13.xx drivers and let 14.xx create them, I got 1 hw error per share completed/block solved. You haven’t completed any, so you have no hw errors.
On a pool where the shares are broken into a lot smaller chunks you would see that. When I wiped out everything and let 14.xx create the bins, 100% hardware errors no matter what settings I used in the pool. I even tried settings like -I 8 -w 8 -g 1 and although was only like 8 kh/s, it still got 100% errors on the pool using the bin created in 14.xx. The only way to get rid of the hw errors was to bring back the bin files from the 13.xx drivers.
I thought the same thing when I dumped the 13.xx drivers and was getting 135 kh/s… No immediate hw errors, cranked along showing new blocks, but when it sat for hours and hours mining, every block it should have solved became a hw error instead. So here, while I should have solved 9 blocks, I got 9 hw errors. https://forum.feathercoin.com/index.php?/topic/7929-gpu-comparison-neoscrypt/#entry68698
-
You have no accepted work…its not running. You had HW errors to start and then they stopped hashing.
Incorrect… Each hardware error was a solved block, it did not stop hashing.
-
I suspect that with that on a pool, you’d get 100% hw errors… When I dumped the bins from the 13.xx drivers and let 14.xx create them, I got 1 hw error per share completed/block solved. You haven’t completed any, so you have no hw errors.
On a pool where the shares are broken into a lot smaller chunks you would see that. When I wiped out everything and let 14.xx create the bins, 100% hardware errors no matter what settings I used in the pool. I even tried settings like -I 8 -w 8 -g 1 and although was only like 8 kh/s, it still got 100% errors on the pool using the bin created in 14.xx. The only way to get rid of the hw errors was to bring back the bin files from the 13.xx drivers.
I thought the same thing when I dumped the 13.xx drivers and was getting 135 kh/s… No immediate hw errors, cranked along showing new blocks, but when it sat for hours and hours mining, every block it should have solved became a hw error instead. So here, while I should have solved 9 blocks, I got 9 hw errors. https://forum.feathercoin.com/index.php?/topic/7929-gpu-comparison-neoscrypt/#entry68698
It has ran 6 hours “the rig in question” and I will be checking it in a short while “its in my shop” I’m still in the house right now and just woke after only 4 hours of sleep.
I’ll post new info and screen shots if there is something new to report and I will also test it on a p2pool once I’ve seen how it has done solo mining.
I hope at least one of the 9 solved blocks my wallet shows were complete on that rig and then and only then will I be able to move forward with my testing for a pool.
-
It has ran 6 hours “the rig in question” and I will be checking it in a short while “its in my shop” I’m still in the house right now and just woke after only 4 hours of sleep.
I’ll post new info and screen shots if there is something new to report and I will also test it on a p2pool once I’ve seen how it has done solo mining.
I hope at least one of the 9 solved blocks my wallet shows were complete on that rig and then and only then will I be able to move forward with my testing for a pool.
Here are the results and also showing hardware errors, now to test a pool.
Sad part was that with the 13.12 bin & 14.9 driver I saw no improvement with my testing & hash rate but I’ll retest to make sure.
-
I’m trying to build a support section with ‘best answers’ to help everyone, could all new issues please be raised in ‘support’
Could we please keep this thread on topic with comparison details only. :)
I’m attempting to split threads out into ‘support section’, so if your thread was in here and it’s missing, please check back there.
-
Here are the results and also showing hardware errors, now to test a pool.
Sad part was that with the 13.12 bin & 14.9 driver I saw no improvement with my testing & hash rate but I’ll retest to make sure.
After testing, there was varied amount of difference of speed between the the 13.xx and 14.xx drivers, using the 13.xx bins. Some were no change, others were 30% or more.
Using the 14.xx drivers with 14.xx bins appears to give across the board improvement in hash rate, but as proven so far they are 100% hw errors.
-
After doing a DDU clean of drivers a 2nd time then installing 13.12 with 13.12.bin I’m getting about the same
hash rate as I was with 14.9 driver and 14.9.bin with the exception of no HW errors, but as seen in the pic I
think the rejects are up a good bit more than they were before I started all this driver changing around and
playing with different bin files. The reason the kh/s doesn’t match is I down clocked my cards a little from the OC
I was using, once those clock are back to the OC they where in my other testing I’m almost 100% sure they will match.
Still using I 14 -w 32 -g 2 as it still seems to work best here for me and my setup so far.
I have 21 different 13.12 bin files of tested settings I will use later when I have more time with 14.9 driver and see
how that goes once its upgrade to.
-
Keep trying, you can get the reject rate WAY down… (I had 58 bin files)
-
AMD Radeon HD 7850,hashrate=40K
cgminer --neoscrypt -I 13 -w 32 -g 1AMD Radeon HD 7970,hashrate=70K
cgminer --neoscrypt -I 14 -w 32 -g 1AMD Radeon HD 6870 ,hashrate
-
GTX 580 hashrate 54 khs
-I 12 -w64 -g 2
GTX580 hasrate 54 khs
-I 12 -w64 -g 2
12 was the sweet spot for intensity
mining pool us.mine-ftc.co.uk
-
Sapphire R9 280x VaporX (Hynix)
14.xx driver with 13.xx bin files
1500/1100 clocks
-I 15 -w 32 -g 2
96-98Kh/s
Getting some rejects but no HW errors !
-
NVidia 750ti
-w 128 -l 15 -g 1 -I 17
seems to be the best I can get as I cant overclock these 3 bad boys on linux
getting around 40khs on main card and slightly less on the two plugged into
pci-3 x1 slots.
-
Sapphire R9 280x VaporX (Hynix)
14.xx driver with 13.xx bin files
1500/1100 clocks
-I 15 -w 32 -g 2
96-98Kh/s
Getting some rejects but no HW errors !
Clocks on 1500/1080 now, no change to hashrate :)
-
Clocks on 1500/1080 now, no change to hashrate :)
Sapphire Vapor-X Tri-X R9 280X with Hynix memory doing 255kh/s - 265kh/s
Intensity 12, worksize 128, 2 GPU threads - clocks are 1150/1500.
Kernel is further modified; compiled on 14.9.
Edit on 11/04/2014:
Same card, doing 290kh/s or so
Intensity 16, worksize 64, 1 GPU thread - clocks are 1160/1500.
Kernel is even further modified; compiled on 14.9
Edit 2, same day:
It does just over 300kh/s on intensity 17.
-
cgminer 3.7.7b
9600GT
Stock Clock-
3.7 KH/S -I 10 -w 32 (66 degrees)
OC (725 Main, 2050 Mem)
- 4.5 KH/s -I 10 -w 32 (69 Degrees - not pushing it hard, old card)
Hash falls as mem clock falls.
Had it laying around, cheers!
-
NVidia 750ti
-w 128 -l 15 -g 1 -I 17
seems to be the best I can get as I cant overclock these 3 bad boys on linux
getting around 40khs on main card and slightly less on the two plugged into
pci-3 x1 slots.
I OC’d each of my 750Ti cards on linux, and upped the TDP to 75W - flash 'em.
-
Sapphire Vapor-X Tri-X R9 280X with Hynix memory doing 255kh/s - 265kh/s
Intensity 12, worksize 128, 2 GPU threads - clocks are 1150/1500.
Kernel is further modified; compiled on 14.9.
Edit on 11/04/2014:
Same card, doing 290kh/s or so
Intensity 16, worksize 64, 1 GPU thread - clocks are 1160/1500.
Kernel is even further modified; compiled on 14.9
WOOOW I want it !!! :)
-
WOOOW I want it !!! :)
I know, the difference between Hynix and Elpida is night and day - my 290X, which I’m not certain, but if I guess from the hashrate, it’s Elpida, does maybe 310kh/s - 315kh/s.
The Elpida 280X, clocked at the same speeds as the Hynix one (1175/1500) gets 240kh/s. So, doing the math, a Hynix 290X would be…
310 + 310 * (290 - 240) / 240 or 375kh/s or so.
