Miners might target Nvidia Maxwell next

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

f1sherman

Platinum Member
Apr 5, 2011
2,243
1
0
Your numbers are indeed realistic , for thoses who say otrherwise
they can check at hardware.fr the cards comsumption measured
in isolation.

http://www.hardware.fr/news/13568/nvidia-lance-geforce-gtx-750-ti-750-maxwell.html

IMG0044029.png


I don't understand.
I can clearly see Sapphire R7 270 Boost:945 MHz, using 111 Watt in Anno 2070. While 750 Ti is using 64 Watts.
How is that comparable to his undervolted 7870 950MHz doing 70W while mining? Or 140 Watt 750 Ti?

This 140w peaks don't reflects on anything but dimensioning the PSU for the card. The card will never draw this 140w of the PSU. Look at r9 290x results on the same test:

Wait a sec, why you are telling me that?
I'm not unhappy with Tom's 170 seconds Metro LL power draw measurement yielding 64 Watt avg for 750 Ti, and the one saying that
the card (750 Ti) takes up to 140W while gaming.
Often taking 100Watts and more.
without appreciating Tom's measurement procedure that is using as tiny as 1ms fluctuations. Because

  1. Tom's time snipets of 1ms are anything but meaningful for our thread
  2. 1ms time window is so short that those 140 Watt readings, can not even be called peaks; not in the sense that peaks usually mean in reviews; for example in in Techpowerup, Xbitlabs or Hardware.fr power draw tests.
  3. There is nothing wrong with noting 1ms power peaks, but know your time window and which one is significant and meaningful for the topic and which one isn't.
    I'd go on the limb and say 1ms time bursts are not meaningful even when selecting PSU, but say 1/4 seconds proly are
  4. We can clearly see 750 Ti using 15 Watts more often than using 140 Watts, and using 35 Watts more often than using 100 Watts.
Shall we use Erenhard's TDP logic and say that 750 Ti is often using 35 Watts and less?

Finally, using his AIDA power measurement methodology on 3 minutes Metro: Last Light, my reference R9 290 is expending 135 Watts at average. I kid you not!

Compare that to Tom's 200+Watt and 150+ Watt minimum for R9 290 in Metro LL.
No wonder his undervolted 7870 achieves only 70 Watts.
 

MrK6

Diamond Member
Aug 9, 2004
4,458
4
81
Why the heck is this so complicated? Plug in your Kill-a-watt, start mining, take the reading. Mining draws significantly less power than gaming.
 

wand3r3r

Diamond Member
May 16, 2008
3,180
0
0
No doubt, idle reading followed by mining.

Surely there are two people here, one with a 750 ti and another with the 7870 with similar settings as described.

I don't get what all the pointless bickering is about. It's easily proven or dis-proven.
 

el etro

Golden Member
Jul 21, 2013
1,584
14
81
Why the heck is this so complicated?

Because you can interpret the results or get the benchmark numbers at the manner you like to get different conclusions. That's why many people doubt about review sites results.

I don't understand.
I can clearly see Sapphire R7 270 Boost:945 MHz, using 111 Watt in Anno 2070. While 750 Ti is using 64 Watts.
How is that comparable to his undervolted 7870 950MHz doing 70W while mining? Or 140 Watt 750 Ti?

His undervolted card not so much(you can do it on a GTX 750Ti too), but the ExtremeTech results shows R9 270(remember that is a underclocked version of the most efficient SKU that AMD sells) mining at the same Hash/w rate that 750Ti haves.

No wonder his undervolted 7870 achieves only 70 Watts.

Its tottally possible, but difficult. May be a binned card...

Shall we use Erenhard's TDP logic and say that 750 Ti is often using 35 Watts and less?

Its a big misinterpretation of your part here. And i don't know what is his intentions with this, but test he put here shows that 750Ti have overriding its 60w TDP(that is not really TDP as the sites claims, but is average gaming power consumption) even on gaming conditions.

Guru3D states ~76 watts as the TDP of 750Ti.
Guru3D states ~90 watts as the TDP of Oced 750Ti.

Finally, using his AIDA power measurement methodology on 3 minutes Metro: Last Light, my reference R9 290 is expending 135 Watts at average. I kid you not!

I Think Aida don't read all the rails that card are being powered(The card have 3 rails powering it at the same time: the PSU 8 pin connector, the PSU 6 pin connector and the PCI-express conector), but anyway, i think the real result of his 7870 is not so bigger compared as he shows. He can simply stress the CPU without the GPU at wprime and then stress the GPU with mining to calculate the difference in power consumption. After this he will end with a card with a better(but not so much) mining efficiency than the original 7870 or r9 270.
 

el etro

Golden Member
Jul 21, 2013
1,584
14
81
More one thing: i think i know what's behind the low Hashrate of 290's: low fan speed.



 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Fisherman. You're 100% correct. However, you're arguing with trolls who are trying to bait you and enjoying themselves too much. Everyone knows mining pushes cards to the max. Generally speaking, a mining GPU load is similar to 100% gaming loads or furmark loads. Actually, mining is usually worse than gaming loads because gaming loads have some variance in power draw. Mining / hashing is straight up 100% pushing the card and stressing it out. It isn't a situation where a card while hashing is going to operate at half power....the cards are pushed to the limit while mining. And I do mean. MAXIMUM. A card mining is AT THE MAX.

Of course everyone that isn't insane knows this. I mean I just had to do a double take. Someone linked an idle power consumption chart in this discussion on mining power draw. IDLE POWER DRAW. Really. Unbelievable. That really happened, and I had to do a double take. It isn't worth arguing against trolls who are baiting you fisherman. They're having too much fun. Not worth it. Pretty sure they know you're correct. But they want to have their fun and probably trying to get you to say something you shouldn't.
 
Last edited:

Grooveriding

Diamond Member
Dec 25, 2008
9,147
1,330
126
Pretty obvious some of you have never done any mining. Mining does not always run your card at its max. It depends entirely on what you're mining. Some coin's algorithms are more intense than others.

Anyways, 750ti has good perf/w for mining, but is still a crap mining card because it's too weak and saving a dollar a week on power is no substitute for mining on a much more powerful card. We'll have to wait and see what happens when the high performance Maxwell cards arrive.
 

f1sherman

Platinum Member
Apr 5, 2011
2,243
1
0
@blackened23
Yeah and I'm like OMG someone is wrong on the internet D:

Now we have the argument of some coins being less intensive. ^^

It seems that now we are talking some exotic coins.
Erenhardt didnt start this with his 7870 mining DOGE? 95% of coins are not Scrypt?
Some games are not less intensive?

Keep them coming you guys, I'm out, there's only so much half truths I can correct.
 
Last edited:

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
but is still a crap mining card because it's too weak and saving a dollar a week on power is no substitute for mining on a much more powerful card. We'll have to wait and see what happens when the high performance Maxwell cards arrive.

We'll have to wait and see. I seem to have heard this somewhere, a lot, before among certain fans.

Wait. And.

See. Yup.

A 60W TDP card doing 260KH at stock and 300KH with an overclock isn't indicative of things to come. For real? What's the hashrate on the 65W TDP R7-250? Someone please answer this. I'm eager for someone to tell us all what the 65W TDP R7-250 does as a hashrate. Am I understanding things correctly. Oh well. I was just told we'll just have to wait and see. Seems to be a recurring theme. It's deja vu all over again. Anyway, it isn't worth it, so i'm going to leave it alone, heh.
 
Last edited:

Cloudfire777

Golden Member
Mar 24, 2013
1,787
95
91
For this test, we only measured the power consumption of the graphics card via the PCI-Express power connector(s) and PCI-Express bus slot. A Keithley Integra 2700 digital multimeter with 6.5-digit resolution was used for all measurements
Maximum: Furmark Stability Test at 1280x1024, 0xAA. This results in a very high no-game power-consumption that can typically be reached only with stress-testing applications. The card was left running the stress test until power draw converged to a stable value.
power_maximum.gif
 

el etro

Golden Member
Jul 21, 2013
1,584
14
81

144 watts on GPGPU with stock volts and stock clocks, temps unknown, fan default.

peaktominning.png


Same 144 watts here, probably at the same conditions(because both sites test with default fan profiles).

Generally speaking, a mining GPU load is similar to 100% gaming loads or furmark loads.

Mining power consumption is close to Furmark power consumption when no cards have Furmark protection.

A 60W TDP card doing 260KH at stock and 300KH with an overclock isn't indicative of things to come.

The 750Ti TDP is not 60W. 60W is the average power consumption of the card, like tom's test said. TDP is more than 60W.

For real? What's the hashrate on the 65W TDP R7-250? Someone please answer this. I'm eager for someone to tell us all what the 65W TDP R7-250 does as a hashrate.

Nvidia not rated the TDP for 750Ti this time.... Toms test show Gaming power cons is ~60w, guru 3d shows max power cons is ~75w(like 7750 and GTX 650)

R7 250 can't catch 750Ti on Hashrate/w, only R9 270/HD 7870/HD 7790 can catch it.




Anandtech tests indicate Furmark protection on Nvidia card:

61475.png





61476.png


Don't know the factor or the ratio of power consumption reduction, but these two graphs shows it is real.
 

ocre

Golden Member
Dec 26, 2008
1,594
7
81
@blackened23
Yeah and I'm like OMG someone is wrong on the internet D:

Now we have the argument of some coins being less intensive. ^^

It seems that now we are talking some exotic coins.
Erenhardt didnt start this with his 7870 mining DOGE? 95% of coins are not Scrypt?
Some games are not less intensive?

Keep them coming you guys, I'm out, there's only so much half truths I can correct.

Yeah, i felt the same way. Glad you stepped up and tried to help. But really, considering the level of insane this has become- it seems pointless.

Like post 80 explains why your results in AIDA are so off for your 290 but still claims the reading for 7870 is to be trusted.....right......
...
illogical. But apparently that doesnt matter.
Hopeless at this level
 

el etro

Golden Member
Jul 21, 2013
1,584
14
81
About the PCI 2.0/3.0 watt draw from PCI connector, many guys over the internet tells that these two standards allow 150w the be draw from the connector only, but no manufacturer relies on this to make their cards.

Here in a Nvidia forums people talking about this.

r9 290x, GTX 690 and Mars 760 overriding their TDP on Furmark:

power_maximum.gif


One more thing to know is that GPGPU is only performance-viable on PCI-ex 3.0. Maybe only in this spec they have enabled the 150w power over the PCI-E slot...

All of this makes(if i am right) PCI-E capable of driving 450w of power to cards. That's why no PC exploded when a 6990 is set to uber mode.

@blackened23
Yeah and I'm like OMG someone is wrong on the internet D:

Now we have the argument of some coins being less intensive. ^^

It more or less clear what i said in my first paragraph in post #80.

It seems that now we are talking some exotic coins.
Erenhardt didnt start this with his 7870 mining DOGE? 95% of coins are not Scrypt?
Some games are not less intensive?

Keep them coming you guys, I'm out, there's only so much half truths I can correct.

I'm not saying I believe it(in the 70w HD 7870 mining)... I'm saying is possible.

Pretty obvious some of you have never done any mining. Mining does not always run your card at its max. It depends entirely on what you're mining. Some coin's algorithms are more intense than others.

We need a mining test suite? o_O
 
Last edited:

Grooveriding

Diamond Member
Dec 25, 2008
9,147
1,330
126
We'll have to wait and see. I seem to have heard this somewhere, a lot, before among certain fans.

Wait. And.

See. Yup.

A 60W TDP card doing 260KH at stock and 300KH with an overclock isn't indicative of things to come. For real? What's the hashrate on the 65W TDP R7-250? Someone please answer this. I'm eager for someone to tell us all what the 65W TDP R7-250 does as a hashrate. Am I understanding things correctly. Oh well. I was just told we'll just have to wait and see. Seems to be a recurring theme. It's deja vu all over again. Anyway, it isn't worth it, so i'm going to leave it alone, heh.

Not sure why you're ranting so off the rails in this thread, maybe in reference to some earlier comments, but not really getting why mine sent you right off the rails ?

'Miners might target Nividia Maxwell next' is the title of thread. For someone buying dedicated mining hardware the 750ti is indeed a crap card. It has a low hash rate and takes up a PCIE slot. You can get a single card that is 3 times as fast in that one slot. Miners focus on what delivers maximum profit. The 750ti's excellent perf/w is not near compensation enough for its low has rate when you are paying cheap NA electricity rates. My statement was completely accurate.

As far as waiting to see what happens when the high performance 20nm Maxwell cards come out; I guess I was not aware we have a seer amongst us who can see into the future. So by all means, because I am really curious to know, tell everyone what card models will be available with high performance 20nm Maxwell, how much they'll cost, what hash rates they'll have, what their perf/w will be and what 20nm AMD cards will be out and the same metrics for those cards.

We'd all love to know because apparently we don't need to wait and see because you already know. Please share.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,600
6,084
136
750 Ti looks to have good performance a watt even compared to an optimized and undervolted 290 (which can be north of 4 kH/W).

However, at current pricing, it doesn't make sense to pick it up when your power rates are low. Not to mention the problem of greatly reduced (1/3) kH/pci-e slot density.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,600
6,084
136
Fisherman. You're 100% correct. However, you're arguing with trolls who are trying to bait you and enjoying themselves too much. Everyone knows mining pushes cards to the max. Generally speaking, a mining GPU load is similar to 100% gaming loads or furmark loads. Actually, mining is usually worse than gaming loads because gaming loads have some variance in power draw. Mining / hashing is straight up 100% pushing the card and stressing it out. It isn't a situation where a card while hashing is going to operate at half power....the cards are pushed to the limit while mining. And I do mean. MAXIMUM. A card mining is AT THE MAX.

Of course everyone that isn't insane knows this. I mean I just had to do a double take. Someone linked an idle power consumption chart in this discussion on mining power draw. IDLE POWER DRAW. Really. Unbelievable. That really happened, and I had to do a double take. It isn't worth arguing against trolls who are baiting you fisherman. They're having too much fun. Not worth it. Pretty sure they know you're correct. But they want to have their fun and probably trying to get you to say something you shouldn't.

^Ignore the parts above about mining. The power draw and component stress (GPU, VRAM, VRM) are all heavily dependent on the algorithm(s) used in a particular cryptocurrency. For something SHA1* based or keccak like Maxcoin it will be considerably less power draw and thermal stress compared to something like scrypt, which in turn is different from something like scrypt-jane.

Even variances in the performance of your cooling will impact power draw significantly. Ever since I put my cards under water cooling, average power draw for a 290 under scrypt mining (one of the most intense) is approximately 200W per card (add another 15W per card for rad fans + pump).
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
I don't understand.
I can clearly see Sapphire R7 270 Boost:945 MHz, using 111 Watt in Anno 2070. While 750 Ti is using 64 Watts.

How is that comparable to his undervolted 7870 950MHz doing 70W while mining? Or 140 Watt 750 Ti?

With a supply voltage reduced to 1V from 1.2V you ll get 44%
lower TDP , that is 77W in Anno 2070, and even less given
the lower temperature of the chip wich will reduce leakages,
so the 70W he quotes for hashing are perfectly possible.
 

ocre

Golden Member
Dec 26, 2008
1,594
7
81
Whatever, its really far fetched.

I dont even care about mining, just cant stand nonsense claims. Seems like people line up to support any far fetched things these days. But can we get back on topic.

As i said, i could care less about mining. But if nvidia cards do become sought after for mining, this would be a win for the gamer. Pricing is out of control for amd cards and this really isnt something that anyone should want. If nvidia did become a choice for just half of the miners, it would make a big impact on prices
 

ICDP

Senior member
Nov 15, 2012
707
0
0
Whatever, its really far fetched.

I dont even care about mining, just cant stand nonsense claims. Seems like people line up to support any far fetched things these days. But can we get back on topic.

As i said, i could care less about mining. But if nvidia cards do become sought after for mining, this would be a win for the gamer. Pricing is out of control for amd cards and this really isnt something that anyone should want. If nvidia did become a choice for just half of the miners, it would make a big impact on prices

If Nvidia became good at mining prices would go up across the board, they most certainly would not go down. If miners are purchasing all the cards supply dries up and prices skyrocket.

This would not stabilise prices, it would simply mean we pay more for less from both Nvidia and AMD.
 

DooKey

Golden Member
Nov 9, 2005
1,811
458
136
If Nvidia became good at mining prices would go up across the board, they most certainly would not go down. If miners are purchasing all the cards supply dries up and prices skyrocket.

This would not stabilise prices, it would simply mean we pay more for less from both Nvidia and AMD.

I disagree. AMD prices are high because they are short on components and supply of their cards is low. Once AMD gets back on an even keel supply should increase and prices will drop. If NV enters the race then price pressure will further reduce due to even more supply on the market. Simple supply and demand.
 

ICDP

Senior member
Nov 15, 2012
707
0
0
I disagree. AMD prices are high because they are short on components and supply of their cards is low. Once AMD gets back on an even keel supply should increase and prices will drop. If NV enters the race then price pressure will further reduce due to even more supply on the market. Simple supply and demand.

AMD are short of components in part (not the sole factor) because of the increased number of cards sold to miners. I did state the word IF as a qualifier.

So IF miners are purchasing a significant percentage of cards for mining, there are less for gamers and prices shoot up because of supply issues. From a reliable source here in the UK, AMD sales have tripled with individuals purchasing hundreds of cards at a time for mining. For the record, AMD cards are NOT in short supply here in the UK and are available for slightly less than MSRP. So clearly the supply issues are only impacting the US market.

Let me put it another way. If it wasn't for miners purchasing more cards, AMD would not have such serious supply issues and prices would be at MSRP in the US.
 

Keysplayr

Elite Member
Jan 16, 2003
21,219
55
91
What about the decrease in AMD sales to gamers to go along with the increase in sales to miners? Surely one would have to agree that the price gouging of AMD cards serves as quite the deterrent for potential AMD buyers/gamers?