Topic: 840 MHash on 5970 - Good or Bad? (Read 5274 times)

Overvolting is essential in the winter. I made enough extra from overvolting to pay for extra cards. The only people not overvolting have high power prices or are scared of damage something. Not that it really matters, next winter gpu's may not be profitable.

There is your problem.

OVERVOLT with 5970 = facepalm.

With those cards you shouldn't even ATTEMPT to overclock / overvolt because the cooing system is already on the edge at stock speeds.

I am confident my 5870s kept below 70 at all times and VRM below 82 at all times and with proper RAM cooling will have no problem passing Furmark or something.

Just make sure the fans are OK though !

I don't use linux, I was unable to overvolt 5970's with it. Running benchmark tests is the best way to know if its working. A lot of cards seem work great for mining but don't work as main displays anymore.

Great. You just made me realize this. Hopefully all my cards are good gamers and working 100%.

How do you recommend testing for gaming worthiness in Linux only ?

I had mine overclocked above 800mh/s in the winter with my pc case in the window. Once the first hot day came around I burned out 2 gpu's. I'm still waiting on diamond to mail them back. Now I try to keep gpu temps at 60-65. A new gpu can run at 70-80c for a while but its not worth it. Especially if ASIC's are coming out. If the video card is not fit for gaming it will be worthless at the end of the year.

It's like a sauna in there during the day...  

I'm quite sure it's more humid than it is outside... oppressive as all hell...  though that could be from different reasons, I guess the water vapor would tend to sink?  My vent is the top of the window (*opens two ways), w/ air blowing in from bottom.

I haven't done the math, but I would imagine that extra water vapour in the air would increase it's heat capacity. Humid air might be able to absorb more heat per volume than dry air of the same temperature, but even if that happened I would imagine the effect would be pretty small. I'd love to know the answer if someone wants to do the math, though.

i have 5 in a room where ambient is outside temperature + a few degrees, so right now that'd be around 95oF, close to 100.  35 - 37oC...  I dunno if heat index matters, since cards don't sweat but who knows. All that humidity in the air might make it harder to dissipate heat from the heat sinks or something, I'm sure someone else here would know.

but.... anyway, I have one system that's water-cooled with 3x 5970... but another with two reference 5970's w/ stock fans.  I run those two at 810 core, 150 memory, @ 1.010v (though I think everything between 1.005 and 1.035 is all the same)...  anyway...  the hottest they get is around 72-73oC, and that's getting about 740mhash or so.   The next step down would be 730/150 @ 0.990v, I believe that's around 660mhash, and drops the temps about 5oC.       oh, this is all with a 25w or so honeywell fan blowing on them

anyway, i guess maybe 1.05v isn't overclocking it.  *boggle*  It's just the highest default performance level, at least on my cards.  i've never taken my 5970's past 1.050v.   i didn't even know they cost go past 1.050v until recently, tbh.  they've always been running on ubuntu.  

and if you run SDK 2.1, then you only lose about 1% performance by dropping memory from 300 to 150 and using worksize 128 instead of 256.

how can i read the VRM temps under linux (BAMT)?

I ran at 860mh/s for a while, 910/300 and 1.1v but the VRM's hitting 120*C made the voltages reset after a couple of hours, so i had to wind back a little bit, but i always push the cards as hard as they can take, i've only ever had fans fail.

Wow what card are you using and what kind of cooling? Is it stable?

My 5970's can only reach 720 tops... Stable at 700MH/s

sergio: tht's true for completely stock card, but you can improve it by rather small investment. I've changed fan (the old one was rattling, the new from eBay is even more efficient) and thermal compound to coolaboratory liquid and both core temps keeps about 65°C under the full load (both cores @ 690/160/0.95). the stock cooler is really great, it just need accompany with equally good accessories (the stock compound seems to me like the cheapest one they could possibly find)

with such a high temperature you risk burning the core that is at the rear of the card since that one gets hotter then the first one in the case of the 5970s.
Anything over 100C you risk burning the card.
I would not ran the card above 95C.

The 5970s can be overclocked safely only if you have it in  a room with cool air.
It is possible to get over 900 Mhash/s with water cooling, however very large radiators will be needed to keep the water cool.

In a hot room  and using air cooling, the card will get about 640 Mhz since it will not be possible to overclock without risk of overheating it.

The diff in temperate between the cores is about 14C, which means the outer core has a much higher risk of overheating.

I the case of the hd5970 it can be highly overclocked using water cooling, it is probably the fastest video card on the market when using water cooling, however
water cooling does not come cheap, be ready to expend over $400 for a good water cooling system. $150 water block video card, $100 water pump, $150 radiator, $50 fittings, heat paste, hoses, fans, heat pads, etc ($50 water temp monitoring), cpu block $50 to $100, etc.

With water cooling the temperatures drop between 60 to 70C and can be dropped even more when using very large radiators, which aircooling at a room at 25C the hottest core will be at about 80C without overclocking, which leaves a little room for overclocking, however at a room temperature of 40C there is absolutely no room to overclock since at stock speeds you will be reaching 100C which can destroy your card.

with water cooling to can ran the card over 50% stock speed, as long as water is kept cool, you do that with air cooling and the card will be destroyed. some may argue that it has a safety feature that when overheated the card automatically shuts down and it is true, however every time the card shutdowns due to an overheat you are taking a high risk, after so many emergency shutdowns one will be a permanent shutdown, and that is a risk you have to avoid.

Yeah probably, I'm not sure how many watts I'm pulling from wall.

This setup is really pretty much defensive (I've tried 680MHz for couple days 24/7 without problems with crashing or overheating) to counter the accumulation of heat in my closed case (which is essential for my rig because of dirty environment).

Luckily I'm currently on flat rate so I don't have to care about MH/W

You would probably get more hashes overall by just lowering clocks and not restarting every x minutes.....  and not just heat and clocks...  electricity.   I get about 2.5Mh/w   above my clocks, you get about 1Mh/w

Yeah, the lower mem probably helps. It is supposed to add few MHashes as well (according to one post I'm not able to find to refer).
My specs are W7 32bit, Catalyst 11.12, Stream SDK 2.1, cgminer 2.4.2

I don't know about long term stability. As a precaution of unstable settings and overheating I've written smal Powershell script which restarts cgminer every X minutes with Y minutes pause between cycles to cool down a bit (currently set to X=240 and Y=10 so I'm actually mining 23/7).

I've tried to go for 700, but that was too unstable (typically crashed in less than 15 minutes).

I completely agree...

you guys can run your cards full throttle if you want to..  I plan on mining bitcoins with my cards for years...  

ajstar:  i run mine at 675.  maybe your lower mem allows the card to go to 690?  how long do they stay stable?

The following rig was restarted May9th. i think because of an upgrade of cgminer .. cant remember.

O, and by the way, this rig only pulls about 850watts

About VRMs: I'm running my 5970 @ 690/160/0.95V/70% with VRMs @ 80°C which I believe is quite reasonable temperature for 24/7 mining. Yes, they are rated for 120°C, but they can operate at that temperature for a limited amount of time. One month of 24/7 mining is equal like a half a year of quite hardcore gaming (4 hours/day). If card is rated to last two years with such a gaming use it means that by mining you will smoke it in less than 4 months.

How many 5970's do you have?

Three of mine are water-cooled and never go above 50oC, four of the GPUs can't reach 850/150, the other two become unstable past 850.

(in reply to: My 5970s run 900 core 300 memory stock voltage. The vrms stay just under 100c and the cores are 55c.  I get about 810 mhash per card. )

Are you talking about this?
That doesn't really sound definitive one way or another.

http://www.xtremesystems.org/forums/showthread.php?242892-EK-5970-vs-stock-cooler-tests/page2

see post 32

he's also posted a ton of stuff on that site, it's very useful.

I wouldn't be comfortable with running anything over 100C. I'd be surprised if any large scale miner allowed their cards to run so hot.

I have an Accelero Xtreme.
100% Fan at all times.
Core temps are around 45C to 50C.
VRM is MAX 110C.

Is that bad?
Or should I try to lower it to ALWAYS be under 100C?

Thanks

Are you sure on that in practice? I've had one where one VRM reached 137 while the others stayed around 100-110C, without throttling the card.

It's not average. If any one hits 125c it will throttle including a couple that don't report to gpuz or other monitors. the guy who designed the 5970 posted on some board a few years back and covered it all I'll try to find his posts.

A word of warning, the 5970s throttle when the average VRM temperature if above 125. If you have two at 110, the other can be at 140 and it will not throttle.

What are your fan speeds? If you're not forcing the fan to a specific level, it will base it on the core temperature with no regard for VRM temps. Better to force the fan a little higher and even if the core is staying cool.

Mine are 1.05v that's "stock". Most 5970 need 1.125 or higher to get over 900mhz. 5870 stock voltage is 1.162

The vrms fry at that voltage the 5870 is 4 phase the 5970 is 3 phase there too much leakage and they overheat.

And learned and the 6990 has the vrm split better and better cooled not sandwiched in the middle.

I have a sapphire card that had 1.18v as the default in it.

What is "stock voltage"?  I have 3 performance levels, the max is 1.05.  I know some BIOSes have 1.10 as "stock", there may be some others even higher.

According to the guy who designed the 5970 the vrms are safe to 150c and throttle over 125c. That's an internal reading unrelated to what you see in gpuz or other monitors however.

The stock heats sinks are far far better than the artics. The TIM they use is the best money can buy.  I have 2 article 5970 xtreme colers in my desk drawer. They are crap for the vrm. The back vrm is adequately cooled and has enough mass and surface area. The front vrm heat sink is a thin plate no surface area or cooling. There a huge empty spot under the front fan that should have been used with a similar setup as the rear or a heat pipe but was poorly designed.

Add to that the fact that if you over tighten the gpu x plates with the artic you'll kill DVI out. The artics are essentially worthless.

I get 10-20 c cooler vrm and within 10 c core with the stock coolers.

My 5970s run 900 core 300 memory stock voltage. The vrms stay just under 100c and the cores are 55c.  I get about 810 mhash per card.

I was just about to say that he might be using AC Xtreme. It tends to keep core temperature very low but the VRMs are way too hot. That's a known problem with that cooler.

45C for core is perfect but 100C for VRMs will fuck them up in the long run. Underclock that rather than overclock.

Well it's running in the garage.
I live in Melbourne and it's about 5C outside. Bloodly freezing but good for the GPUs.
The GPU also has an Artic Accelero Cooler, running at 100% at all times.

Also is that current throttle bad for the life of the GPU?

840mhash seems slow, for 910/375.  Probably using a crappy SDK.

I have no clue as to how you can run it stable at 910/375 @ 1.050v, either.  Only two GPUs out of 10 of mine can run stable at 850/150 @ 1.050v.  The rest are lower.

At 850/150, it gets 800mhash.

(well, I just noticed that your screenshot actually shows that you have increased your voltage past stock, which would explain the 910)

ed:  and your card wouldn't be running at 43.5o, either

It's running at 45C. How is that too hot?
112C is the VRM temp I presume? CGMiner says I'm at 45C, same with MSI AfterBurner, etc. I'm not sure where this 112C temp came from, thus me thinking that it is the VRM?

EDIT:
I've made them be 890 Core, and 375 Memory. 1.050 Voltage. Every thing seems fine now, VRMs seem to be about 105C.

Also I had a look at this, my current on the card sometimes bounces, or throttles, is this bad?

Too hot. I keep my cards under 65 but most of us run them between 70 and 80 I think?

How much do you pay for electricity? Do you have a kill-a-watt?
http://tradebtc.net/bitcalc.php
Try different numbers and see what gives you the best ROI.

I would try settings around 800/150 or 850/150. You have to find the right voltage.

Over 100C is simply too much.

How hot should it be max?
It's been running like that for 2 days straight. Is that bad?

Looks too hot to me.

I'm not sure if that 110 C temp is the VRM?

I'm hoping it is. Because if it's not the VRM is probably smoking hot.

Also what is a safe range for the VRM temp? Under 120 C?

Thanks

Hi I'm relatively new to Bitcoin Mining and was wondering if about 840 mhashes is good for a 5970?

I'm currently at 910 mhz core, and 375 mhz memory. Stock voltage. Cgminer reports voltage as being 1.050, so I believe that is stock? Or is that too high of a voltage?

Also with such a high clock with this hurt my card? Eg will it reduce the life of my card?

Thanks.