AMD: Southern Islands (7*** series) Speculation/ Rumour Thread

At same time, i dont know why they dont overclock the 7970 ... if they want make a 680 OC article, good to put it agains the stock 680, but what is doing the stock 7970 in this article ?


actually i have test my 7970 (second card is in RMA ) with 0% TDP all the round to 1300mhz.... i have start to use +20% TDP limit only after 1350mhz, a run on unigine have fail at 1350mhz . ( who have pass without problem then with +20% ) ..

No throttle at all... I dont know exactly how is working the slider, if it is linked with temp or not .. at 1220mhz, the temp is only of 42°C max.. stock i got 36-37°C full load max.

The temp full load with this waterblock are just completely sick.. never seen so low temperature on a card..

System:
- Laing D5 ( 8-24V ) + EK X-Top rev2 ( absolutely genious, high perf + silent )
- 360mm Rad EK XTX + 3 Gentle typhoon ( 1450 AP14 - 100% silent, way better of my Typhoon 1850rpm )
- EK Bay Spin reservoir
- EK Supreme HF Nickel
- 2x EK 7970 Nickel + Acetal EN. ( second card is in RMA )
- 1/2" White tubing, compression fitting, EK coolant
- Gelid Extreme Thermal grease


3Dmark11, max temp 42°C: 1212mv @ 1221mhz/1500mhz :
(sick temp, cpu is at 5.1ghz in the loop )








1315/1500mhz 1.25V ( not maxed just a default run )




default outside Tesselation normal: ( Exactly half of what i get with 1150mhz CFX )
single HIS 7970 1300/1500mhz



16xAF + Extreme tesselation ( no AA ) 1250mhz


Conclusion, if you are searching a waterblock for your 7970's... get a look on EK waterblock site.
 
Last edited by a moderator:
Yeah, I noticed that too. I was looking for the 7970 to also be overclocked from the get go and it wasn't there. I think we can drop the temps variable from overclock reviews and just make sure that their is no throttling. There is more then the gpu that could be a concern such as mosfets, vregs, vram, etc which may not offer temps should be also be properly cooled by increasing the fan speed. On full block WC solutions that should negate some of that.
 
Good idea Jared , or just do a hardmode. when my second card is back .. note i was just speak about the TDP limit slider. i can use 0% till 1300mhz-1325mhz. then it start to need +20%. (on unigine, i dont speak about a stable 24/24 oc ofc, even if unigine is not what i will call the most tender test ).

I have not test the max OC yet. look like i can run at 1350mhz with 1270-1280mV... so 1375-1400mhz maybe is possible with 1300-1315mV. I dont know exactly what the limit is, and i will not forcibly make suicide run without Dice lol.


Yeah, I noticed that too. I was looking for the 7970 to also be overclocked from the get go and it wasn't there. I think we can drop the temps variable from overclock reviews and just make sure that their is no throttling. There is more then the gpu that could be a concern such as mosfets, vregs, vram, etc which may not offer temps should be also be properly cooled by increasing the fan speed. On full block WC solutions that should negate some of that.

Yes, with a full cover waterblock, the temperature of mossfets, power regulator etc are night and day.. even the top of the card stay cold on this place ( if you have seen a 7970, this place get quickly really really hot )
 
Last edited by a moderator:
Proactive vs Reactive cooling of OC cards...

Yes, with a full cover waterblock, the temperature of mossfets, power regulator etc are night and day.. even the top of the card stay cold on this place ( if you have seen a 7970, this place get quickly really really hot )

You know what would be interesting? To see if there is a performance delta (using a line graph) of an OC @ 1200/1400 using the stock HSF using automatic fan speed vs manual fan speed of 50%, 60%, 70% and 80%. Just to see if the card performs better when you have proactive vs reactive cooling. It shouldn't be limited to just the 7970 either, I would like to see 680 thrown in there as well.
 
This will mainly depend of the card, i have 2 different branded 7970 ( second is in RMA ), the 1st i bought was a Sapphire ( really early model, the first wave ), the second is an HIS, and they absolutely dont act similary on fan speed.

The Sapphire one was upping really fast the fan, when the HIS was stay really cool.. it seems 40% on Sapphire was faster of the HIS one.

Now like the Sapphire is dead, the problem was maybe a bad contact or mosftets problem.

Now this is clear if you let the fan at 60% fixed ( so even when the card is idle ), the temperature will be lower. ( lower temp, less power leaks, so less tdp. ) ( i remember some tests of the GTX580, with different speed, and TDP, the tdp was going down following the temperature )

If i have the time when i got the second card back, i will see if i can test that before i build the waterblock.

Now i think with the Turboost on the 680, this will be really informative, just for check if it is too linked to temperatures. if the card stay at lower temp + lower tdp, the Turbo Boost should be maxed most of the time.
 
That's really impressive even if cherry picked. It just goes to show what the gpu is really capable of. Take a look at the temps further down the page. There is a huge difference in idle and load temps. At 1st I thought that there was some sort of increased fan speed but at only 20% that's not much at all.
 
As I have mentioned a few times in other posts, I have one of those Sapphire 7970 OC boards and it works exactly as is described. I've also had it up to 1.3Ghz GPU speeds at 1.225v although I haven't done extensive stability testing at those speeds (so far, it seems ok.)

It's a fantastic card and VERY quiet.
 
That's very impressive then. I have to ask what gpu chip is being used vs the regular 7970? I don't get the impression it's the same as the stock cards.
 
Dunno. GPU-Z shows it as an '84% ASIC', but Dave has told us that number isn't a pure indicator of anything, really.

Minus taking it apart, is there something you'd like me to check for you? I'm happy to supply screenshots or pictures or whatever.
 
Dunno. GPU-Z shows it as an '84% ASIC', but Dave has told us that number isn't a pure indicator of anything, really.

Minus taking it apart, is there something you'd like me to check for you? I'm happy to supply screenshots or pictures or whatever.

A screenshot of the top of the die might reveal something or be a reference point for others. Perhaps someone might comment on it if there is anything we need to look for. Thanks.

But if I was in the market for a particular 7970 the 7970 OC Edition would be the one I would be looking at.
 
Last edited by a moderator:
That's really impressive even if cherry picked. It just goes to show what the gpu is really capable of. Take a look at the temps further down the page. There is a huge difference in idle and load temps. At 1st I thought that there was some sort of increased fan speed but at only 20% that's not much at all.

That might actually be a big part of the explanation. Leakage is a function of temperature, so this large temperature gap could help the Sapphire model draw less power. Given the small difference in power, it's probable that the lower temperature is the cause of the lower power more than the other way around. And that's thanks to Sapphire's custom cooler, of course.

Throw in the usual variability, and maybe a slightly more power-efficient board design, and there you go.
 
That might actually be a big part of the explanation. Leakage is a function of temperature, so this large temperature gap could help the Sapphire model draw less power. Given the small difference in power, it's probable that the lower temperature is the cause of the lower power more than the other way around. And that's thanks to Sapphire's custom cooler, of course.

Throw in the usual variability, and maybe a slightly more power-efficient board design, and there you go.

It could very well be but I was under the impression that the more efficient a chip is, including lower leakage, the less likely it's a good overclocker.
 
Noticed 4 in stock 7970's SKU's now priced at 529 at newegg...defacto price drop.

also seeing discount throught the line now on a few sku's. 429 7950, 239 7850, even a 329 7870 (surprised as that seems to be the most in demand card)
 
Snapshot of the ASIC quality and a few other pertinent details on the current rig. I was wrong, the quality indicator reads "86.1%" for whatever that's worth.
7970%20Capture.PNG
 
Noticed 4 in stock 7970's SKU's now priced at 529 at newegg...defacto price drop.

also seeing discount throught the line now on a few sku's. 429 7950, 239 7850, even a 329 7870 (surprised as that seems to be the most in demand card)

Heh, one thing I noticed on the 529.99 USD cards. No free shipping on any of them while any cards 549 USD and higher has free shipping. So they are actually 536.56 to 536.87 USD compared to the other cards that are still as suggested MSRP.

Also only the stock clocked cards are dropping in price. All the OC versions aren't budging at all with regards to price. That makes me somewhat wonder if they are trying to clear out stock clocked cards to potentially make room for a new SKU. Probably not, but who knows... And it certainly wouldn't surprise me if we are seeing a slow migration down to the 499 USD price point.

Regards,
SB
 
You can pick up the xfx black edition (1GHz clock) over at ncix for $539. That price supposedly ends today tho.
 
It could very well be but I was under the impression that the more efficient a chip is, including lower leakage, the less likely it's a good overclocker.

I believe that's correct (all other things being equal) but only for the chip's "intrinsic" tendency to leak, so to speak, not temperature-induced leakage.
 
Back
Top