• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Looks like Scrypt mining is fixed in Maxwell

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Does anyone actually mine on these lower end cards? Even at a good performance/watt ratio these are still 'slow' cards for mining.

I haven't mined in a while, been too busy trying to break my FX with crazy overclocks. 😛 But I thought mining was dominated by big GPU's yet.

KH/S per slot is also an important metric.

That's why I think the card, currently, that has the best chance of curbing demand for the 280X and up would be a 7990 or equivalent. Make it liquid cooled and single slot and watch it fly off of the shelf for ~$1K.
 
KH/S per slot is also an important metric.

That's why I think the card, currently, that has the best chance of curbing demand for the 280X and up would be a 7990 or equivalent. Make it liquid cooled and single slot and watch it fly off of the shelf for ~$1K.

I'm kind of surprised AMD has not announced a monster 600 watt dual Hawaii full length card for $1500. Miners wouldn't care that it doesn't fit in most cases - half of them seem to not even use cases.
 
When the rest of Maxwell launch? I still have time to return my R9 290s and get Maxwells instead, but I'm mostly interested in higher-end cards, not 750 Ti's.
 
When the rest of Maxwell launch? I still have time to return my R9 290s and get Maxwells instead, but I'm mostly interested in higher-end cards, not 750 Ti's.

Anytime this year. Your guess is as good as mine. No official ETA other than "2014" from what I have seen.
 
KH/S per slot is also an important metric

Stop it. You're reminding me of the $550 7990 at Amazon before Black Friday. Why didn't I buy 6 of them then? All I would have needed is two very strong PSUs with them, and I would have had a nice 10 MH/s spread over 2 rigs taking up very little room as opposed to the 6 I have right now. 🙁
 
are miners showing any interest in it? It seems that 7 750ti's would roughly equal 2 290x's, with the nvidia solution being cheaper and more power efficient. Though i think you would need an x79 board 7 pci express slots.

edit: well, i looked at the price for asus's x79 board with 7 pci express slots. Damn, $450. Still economically neutral though. 2x 290x=$1400, $1400+$150 mobo is $1550 vs 7x750ti= $1050, $1050 + $450 mobo is $1500.
 
Last edited:
When the rest of Maxwell launch? I still have time to return my R9 290s and get Maxwells instead, but I'm mostly interested in higher-end cards, not 750 Ti's.

Maybe a product to replace the 660 (750 Ti Boost?), but I think that would be it until Maxwell 20 nm. That's not coming out until 3Q at the earliest.
 
http://cryptomining-*********/tag/gtx-750-ti-scrypt-mining/

https://docs.google.com/spreadsheet/lv?key=0AlhSF602y9DSdHRneWVZcjJLWmhjaVBMTHVCR3l4eHc&toomany=true

750ti: 265kh/s at 60w (300kh/s with overclocking)
7750: 180kh/s at 55w

Once the big-core Maxwells arrive they should be hashing monsters. If anyone is wondering what is going to cause a price drop for the R9 290s some competition from nVidia is finally arriving.

Will bring what? Two Vendors with overpriced cards, and we left in the dust.
 
Will bring what? Two Vendors with overpriced cards, and we left in the dust.

I think there would be a bunch of cheap GCN cards on ebay if ppl began replacing them. But yea, all signs point to 20nm maxwell cards being horribly overpriced like current radeon cards.
 
Last edited:
I think there would be a bunch of cheap GCN cards on ebay if ppl began replacing them. But yea, all signs point to 20nm maxwell cards being horribly overpriced like current radeon cards.
I find that miners don't swap out cards in their setup, they simply add to them. Even if Maxwell is a monster, I'd expect that demand for new AMD cards will drop but miners will not sell their existing cards, as long as the operating revenue from the old cards exceed (and even just come close to) their electricity costs.
 
I like how nvidia has 192-bit memory bus on some cards
makes for an interesting price/perf ratio, too bad these first ones are 128-bit cards
 
Until Nvidia lowers their pricing they won't be overtaking radeons for mining.

These days, a R9 280X is over $450. At $150 apiece, you can get three 750 Ti cards for that money. Based on what has been posted so far, the Nvidia cards would have a total hash rate of 795 KH/sec (265 x 3) at stock, and if overclocked, could go up to 900 (300 x 3). That's slightly better than you would get out of the AMD card.

Of course, the Nvidia cards would require more slots, which means you'd probably have to get some powered risers. That adds to the cost a bit. On the other hand, the Nvidia cards will use less electricity.

If AMD could keep to MSRP, they'd still be winning the cryptocurrency game by a sizable margin. But they can't seem to provide adequate supply at this time.
 
For big scale mining, indeed Hash per slot is the primary concern, efficiency matters but not as much for people who pay <5 cents a kWh compared to overall hash power and /slot.

I pay 30 cents a kWh so to me, efficiency indeed matters.
 
are miners showing any interest in it? It seems that 7 750ti's would roughly equal 2 290x's, with the nvidia solution being cheaper and more power efficient. Though i think you would need an x79 board 7 pci express slots.

edit: well, i looked at the price for asus's x79 board with 7 pci express slots. Damn, $450. Still economically neutral though. 2x 290x=$1400, $1400+$150 mobo is $1550 vs 7x750ti= $1050, $1050 + $450 mobo is $1500.

Not really economically neutral considering the fact that if 2x290x takes up only two slots, so even with a $150 mobo, you could still add at least 3 more cards with powered risers. With the 750Ti config, you would have to build another rig to expand your hashrate which pretty much defeats the purpose of going that route.
 
The fact of the matter is that the 750 ti is the new king of hash/watt, but not hash/slot.

This means it is indeed the best pick for certain limited situations. For example, consider the typical enthusiast who has at least 2 functional PCs worth of "old" parts laying around. These would have been limited previously by their older, generally smaller power supplies (I assume most unused power supplies are unused by fact of being too small or having insufficient connections). At only 60 watts per card, even older C2D era stuff with 2 pci-e slots and a modest 350-400w power supply should be able to easily handle 2 750 Ti's hashing away merrily.

Obviously this isn't scalable and depends on each person. But I personally have at least 2 computers worth of old parts I'm strongly considering turning into miners. I have a couple rooms in the house where the heating doesn't reach as well and I'd just stick them there as space heaters. Even going back to C2D, the downclocking and low energy states on the CPUs weren't terrible. I'd estimate around ~220-240 watts constant for such a machine at the wall.
 
For big scale mining, indeed Hash per slot is the primary concern, efficiency matters but not as much for people who pay <5 cents a kWh compared to overall hash power and /slot.

I pay 30 cents a kWh so to me, efficiency indeed matters.

Hash per watt matters once people hit their max power limit.

However I doubt people will switch until GM206 or GM204 comes out, since by GM206 the card will be about equal to a 7950 in hashing power (If it also is on TSMC 28HP, much higher if it's on TSMC 20SOC) and that's the lower preference limit of people who care at all about hash/slot.
 
Last edited:
These guys measured hash per watt
LiteCoinEfficiency.png

IF AMD made any moves to improve their integer performance on 20nm plus the additional functional units that node shrinks afford, they might just hold nvidia off
 
Interesting but what will be really interesting is high end 20nm vs. 20nm. They are so weak that they are hardly worth considering unless you are extremely concerned with perf/w, in which case it's only equal to the 270, while being slower.
 
Back
Top