I don't get it. A GPU is just a vector co-processor that can run any compatible software application.
Different architectures perforrm differently for different applications, just like a CPU. Mining is not
an singular application but a group of applications (kernels), each with their own performance
profile.

So what is a mining GPU? Does it have more cores to improve compute-bound applications, or a faster
memory interface to improve I/O bound applications? It depends on the algo.

So what features could their be that define a GPU as a mining GPU? Just removing the video ports won't
reduce the cost enough to draw miners, or anyone. to buy it instead of a regular GPU.

I suspect it's just going to be a marketing campaign to miners. They may brand a card as mining vs gaming
but it will only be cosmetic.

Haha. it only became an issue about a typo when that was the only way he could win the argument. So technically he did win.
He only claimed that "Myr-gr" the text was removed from the post, not the algo from the miner as I foolishly assumed.
I got trolled into a pointless argument

But the issue of confusion among the groestl algos is real.

Groestl and Diamond-Groestl use the same hashing functions but different merkle root encoding. They are not interchangeable
but produce the same hash rates. Myriad-Groestl uses diferent hashing functions and produces a higher hash rate.

My apologies to everyone reading this thread for feeding a troll. I have deleted my posts.

Confusing Myriad-Groestl and Diamond-Groestl? Diamond is the same speed as Groestl, Myriad is faster.

If you're talking about cost of cooling you must be talking about air conditioning. Humidity matters with AC.
With fresh air cooling humidity doesn't matter. Your biggest challenge with fresh air cooling is air circulation.
An ambient tempterature of 32C is ok if you can exhaust the heat well. A couple of box fans is a lot cheaper
than AC.

I haven't done the math but the cost of air conditioning will probably kill any profit potential.

There was another recent thread discussing the effects of humidity.

cpuminer-opt 3.6.5 released with support for jha algo (jackpotcoin, sweepstake) wiith AES optiizations.
Cryptonight should also be a little faster.

jha disappeared from yiimp before I could do a final test, I don't know if it's coming back.

git: https://github.com/JayDDee/cpuminer-opt

tarball: https://drive.google.com/file/d/0B0lVSGQYLJIZYWtaV1YxMV9hWkE/view?usp=sharing

Windows binaries: https://drive.google.com/file/d/0B0lVSGQYLJIZNm55Qjc5emwyU3c/view?usp=sharing

It should be easy but I'm having a weird problem.

The issue with humidity and electronics is condensation, it's not like how it affects warm blooded animals.
The moisture also gets absorbed by some components shortening their life. Low humidity is good but it has
nothing to do with cooling perrformance.

High humidity environments is an ironic case where higher temperatures are better to keep the relative humidity lower
and protect against condensation.

Some "light" reading...

https://www.electronics-cooling.com/2013/06/electronic-performance-impact-of-elevated-humidity-environments-implications-for-free-air-cooling-of-data-centers/

There are a couple of interesting points that I would  like to comment on.

Mining scrypt algo requires ASIC rigs for any chance of making a profit. The ASIC market is competitive with newer
faster ASICs  becomming available limiting the usable life of any ASIC hardware purchase.

Using 5 year old GPUs willl likely not be profitable for any algo. They perform at about 30% of similarly priced current
generation GPUs and consume about twice the power. (made up stats but they're in the ballpark). Software support
is also more challenging. Newer versions may not support older hardware and older versions will be less efficient
(less optimized) and may not support newer algos. Being able to compile the software yourself could alleviate that.

Quantum computing is still just a dream. IBM is starting an "initiative" to build the first commercially available
qauntum computer, I saw no timetable. Quantum computing is still a long way from reality and even longer
beforre it reaches the the consumer market.

How long can mining last? I have been asking myself that question for 3 years and I still don't know the answer.
No one really does.

The size of the buffer is determined by the N factor. I'm not familiar with verium, maybe someone in the verium
thread can help.

Edit: you didn't mention the number of threads, one buffer is allocated for each thread. Try with fewer threads.

You don't have enough memory.

Do more reading before asking stupid questions.

Yes because the algo doesn't support it.

English?

I have to disagree with the suggestion to install 3 GPUs on the mobo.
Most mobos don't have adequate spacing for cooling with 3 GPUs, and cooling is critical
for mining. Two on the mobo and the rest on powered risers connected to x1 slots.
And make sure you use a good PSU with lots of extra power, especially if you intend to overclock.

Git is being a pain in the ass. I still can't upoad the Windows binaries of the v3.6.4 release.
Previously I was gettng every possible error from file too big to file is empty to file already exists
to invalid file type.

Now things are different, it doesn't do anything, no error no upload. Furthermore while playing around
the release got published just by saving it as a draft. And now I can't delete the release, any attempt does
nothing.

The result is the git release feature is not reliable and of no use at this time. Please use Google to download
cpuminer-opt.

Don't use Lightshot, I'm not installing that just to see your screen cap. cpuminer-opt is not a GUI program
so don't show me images. A simple text copy and paste will do.

It is a security red flag to receive a file that requires installing an app to read. DON'T DO IT, EVER!

That's not a coin, it's an algo. Is there a coin that s uses it for POW?

That would assume the software is currently throttling performance, which it isn't.

If it crashes with only one GPU, ie one CPU thread, it's not a thread race condition.