Topic: tom's hardware including Bitcoin Mining in 2012 GPGPU benchmarks (Read 6919 times)
What? Stable algorithm that never changes? One kernel update and the results can be 10MH/s faster or slower. One wrong setting on intensity, worksize, etc and the result can be way off. Also, what happens if during the test the mining server or network in between gets flaky? Then the results aren't stable. I'm all for review sites showing the hashing power of the cards to the public, but it can't be considered a 100% stable, no variables test like they are claiming.
April 11, 2012, 01:18:13 AM
And pray tell, which new site is this?
I dind't say there was a site doing it.But usually demand drives offer, so, if GPU mining will be still relevant in the near future (thing I doubt) I bet someone will pop up with decent and useful benchmarking.
April 10, 2012, 11:02:13 PM
They have 6990 performing worse than 6970. None of the values have any relevence to real world. The amount of error varies from card to card making even relative comparisons invalid.
The "benchmark" is beyond useless and beyond any defense.
The "benchmark" is beyond useless and beyond any defense.
April 10, 2012, 10:49:57 PM
Since their BTC benchmarks are so untrue and it seems they have little will to do anything to inform themselves to correct this, it makes me wonder if indeed all their other benchmarks could be of the same rubbish quality 
That would be a very poor conclusion. Every benchmark shows the result from running a given program. They aren't trying to show what the fastest hash rate is; they're trying to show what the hash rate is for that hardware with that program with no tweaking. It would be a ridiculous waste of time for them to try a bunch of different programs and settings to try to tweak the result for every single piece of hardware that they test. Absolute values aren't important. Comparative values are what is important. And if the programs or drivers are optimized to work better on certain hardware, that would be nothing new or unusual in the benchmarking world. April 10, 2012, 07:54:15 PM
A few reasons I know
1 Mining which involves profit and often recruiting to make it worth it.
1 Mining which involves profit and often recruiting to make it worth it.
Recruiting to make it worth it!? In real terms... recruiting more members will raise the difficulty making it worth less?? The only increase will come from more mainstream use. Idiots. Maybe I'll go post something on their discussion boards as I have a pretty high post count with them.
April 10, 2012, 03:30:48 PM
Of course that is all utterly ridiculous and stupid, but that is the reasoning they use.
That's fine, they'll lose users to a new site that will do serious mining benchmark 
April 10, 2012, 02:25:54 PM
Of course that is all utterly ridiculous and stupid, but that is the reasoning they use.
That's fine, they'll lose users to a new site that will do serious mining benchmark 
April 10, 2012, 01:18:28 PM
Anybody on here also members of Overclockers.com or [H]? Do they have Bitcoin subforums yet? Those guys are big fans of crunching projects so it'd be good to have them on board.
overclockers.com hates bitcoin.Why? Its been a while since I was active on there.
Not sure if there was a mixup or it's just two different sites that both "hate" bitcoins, but I found this at overclock.net forums-
http://www.overclock.net/t/1104285/owai-why-is-bitcoin-discussion-not-allowed
Quote
A few reasons I know
1 Mining which involves profit and often recruiting to make it worth it.
2 It is not an internationally recognised currency
Quote:
6.3. Currency: Only internationally recognized currency may be traded with. Any synthetic form such as 'BitCoin' is NOT recognized here and and not allowed here.
http://www.overclock.net/marketplace...rules-wip.html
I know that is the marketplace rules but
Quote:
These rules supersede all previously posted rules and are part of the Terms of Service of Overclock.net.
We also don't allow any hint of it.
Any posts referencing, hinting at etc should be reported
Any further questons should be directed to Chipp or The_Manual
Now we have that cleared up any further threads/posts by anyone will be punished
1 Mining which involves profit and often recruiting to make it worth it.
2 It is not an internationally recognised currency
Quote:
6.3. Currency: Only internationally recognized currency may be traded with. Any synthetic form such as 'BitCoin' is NOT recognized here and and not allowed here.
http://www.overclock.net/marketplace...rules-wip.html
I know that is the marketplace rules but
Quote:
These rules supersede all previously posted rules and are part of the Terms of Service of Overclock.net.
We also don't allow any hint of it.
Any posts referencing, hinting at etc should be reported
Any further questons should be directed to Chipp or The_Manual
Now we have that cleared up any further threads/posts by anyone will be punished
Of course that is all utterly ridiculous and stupid, but that is the reasoning they use.
April 10, 2012, 04:13:46 AM
Since their BTC benchmarks are so untrue and it seems they have little will to do anything to inform themselves to correct this, it makes me wonder if indeed all their other benchmarks could be of the same rubbish quality
Once I compared TH with some russian hardware reviewers, like overclockers.ru, 3dnews.ru and etc. It largely depends on the author, and all articles are biased to some extent. All I can say is that TH is as much biased, as any other sourse. If you're lucky enough to find a high quality writer (aka more or less objective), just stick to them
April 10, 2012, 03:32:11 AM
Since their BTC benchmarks are so untrue and it seems they have little will to do anything to inform themselves to correct this, it makes me wonder if indeed all their other benchmarks could be of the same rubbish quality
+1 
April 10, 2012, 03:16:16 AM
Since their BTC benchmarks are so untrue and it seems they have little will to do anything to inform themselves to correct this, it makes me wonder if indeed all their other benchmarks could be of the same rubbish quality
April 09, 2012, 07:38:38 PM
Anybody on here also members of Overclockers.com or [H]? Do they have Bitcoin subforums yet? Those guys are big fans of crunching projects so it'd be good to have them on board.
overclockers.com hates bitcoin.Why? Its been a while since I was active on there.
April 09, 2012, 03:57:23 PM
Anybody on here also members of Overclockers.com or [H]? Do they have Bitcoin subforums yet? Those guys are big fans of crunching projects so it'd be good to have them on board.
[H] has a distributed computing sub forum with a thread for Bitcoin. Most of it is F@H however. We are generally more interested in builing 4P systems for F@H on the forum.
XtremeSystems has a bitcoin sub-forum but it receives very little attention due to WCG being the primary project over there.
April 09, 2012, 12:54:23 PM
Anybody on here also members of Overclockers.com or [H]? Do they have Bitcoin subforums yet? Those guys are big fans of crunching projects so it'd be good to have them on board.
[H] has a distributed computing sub forum with a thread for Bitcoin. Most of it is F@H however. We are generally more interested in builing 4P systems for F@H on the forum.
April 09, 2012, 10:44:31 AM
Anybody on here also members of Overclockers.com or [H]? Do they have Bitcoin subforums yet? Those guys are big fans of crunching projects so it'd be good to have them on board.
overclockers.com hates bitcoin. April 07, 2012, 06:58:10 AM
Anybody on here also members of Overclockers.com or [H]? Do they have Bitcoin subforums yet? Those guys are big fans of crunching projects so it'd be good to have them on board.
April 06, 2012, 09:37:47 PM
Doesn't go into operating costs or anything. Also a 6990 is faster than a 6970, lol. I'd take their results with a grain of salt.
April 05, 2012, 12:00:49 AM
yep,i asked in my comment about overclocking and using a better miner and stuff
I think a few emails offering help could be good
I think a few emails offering help could be good
April 04, 2012, 09:43:27 AM
My 7970 gets over 550mhash at 950 core and under-clocked memory. I don't know how they come up with that result.
They have no idea what they're doing. I have to imagine that the 6990 is mining on only one GPU and they figured that XFire would take care of the rest. How else do you explain it getting almost the same but slightly lower performance than a 6970?
Should send them an email letting them know that I would retest all their cards using a couple miners to get the optimal settings for each, if they'd be willing to send me all the cards for a couple months.
April 04, 2012, 09:38:08 AM
My 7970 gets over 550mhash at 950 core and under-clocked memory. I don't know how they come up with that result.
You can see their updated results here http://www.tomshardware.com/charts/2012-vga-gpgpu/14-GPGPU-Bitmining,2970.html
Which still seem too low. I know the 7950 I have gets 414 at stock clocks with cg miner and that's what their 7970 gets.
Which still seem too low. I know the 7950 I have gets 414 at stock clocks with cg miner and that's what their 7970 gets.
2012-04-03
http://www.tomshardware.com/reviews/graphics-card-benchmarks-charts-review,3154-8.html
Awesome!
signed up and commented - did anyone else?http://www.tomshardware.com/reviews/graphics-card-benchmarks-charts-review,3154-8.html
Awesome!
Looks like bitcoin is gonna gain a lot more momentum soon with all these mentions. +1 for the cause!
2012-04-03
http://www.tomshardware.com/reviews/graphics-card-benchmarks-charts-review,3154-8.html
Awesome!
signed up and commented - did anyone else?http://www.tomshardware.com/reviews/graphics-card-benchmarks-charts-review,3154-8.html
Awesome!
Also, if anyone wants to do a straight swap of my 6970 at an impressive 390MH/s for your 6990 at a pathetic 389MH/s, we could work something out.
Yeah it might be too much to ask but it would be nice if they ask for HERE for some assistance of proper mining configurations.
I mean yeah there is some variance depedning on the particular miners and params but 300 MH/s for a 7970 is just a useless datapoint.
I mean yeah there is some variance depedning on the particular miners and params but 300 MH/s for a 7970 is just a useless datapoint.
Are they going to start using correct miners? They were showing the 7970's max mining potential as 300mhash or so for a long time, while you could easily get double that with diablo...
oh that's who mysteryminer2 is!
About that Bitcoin vs. F@H thingie in the comments:
Unlike F@H Bitcoin has a 100% stable algorithm that has to be executed on every worker. F@H or any other BOINC project might update their algorithms over time and invalidating the older benchmarks. Bitcoin mining only might make use of updated OpenCL kernels if someone discovers that there is a more efficient wasy of combining some AND and OR instructions or the like. The algorithm itself ("SHA256(SHA256(header))<=Difficulty") won't change though.
Unlike F@H Bitcoin has a 100% stable algorithm that has to be executed on every worker. F@H or any other BOINC project might update their algorithms over time and invalidating the older benchmarks. Bitcoin mining only might make use of updated OpenCL kernels if someone discovers that there is a more efficient wasy of combining some AND and OR instructions or the like. The algorithm itself ("SHA256(SHA256(header))<=Difficulty") won't change though.
Jump to: