Associate
Not the power limit, the powertune, which is adjusting the card automatically...switching off not used parts for instance
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
The Fury(X) Fiji Owners Thread
- Thread starter Kaapstad
- Start date
Not the power limit, the powertune, which is adjusting the card automatically...switching off not used parts for instance
Techpowerup measured off a coil on the pcb and claims a measured voltage held in a 3d load was 1.22v, so don't think it's drivers at all, more it needs the vrm controller unlocked to suck more current.
Remember I said about my fear over thermal density in particular on air cooling, let's see when voltage is unlocked on the fury-x/ in particular fury.
Last edited:
Caporegime
Just tried some memory overclocking.
Stock bench results
1120/550 (crashed at this speed but tried again and it completed)
Stock bench results
1120/550 (crashed at this speed but tried again and it completed)
Caporegime
Oh, the power saving feature? 0 core and all that....
Techpowerup measured off a coil on the pcb and claims a measured voltage held in a 3d load was 1.22v, so don't think it's drivers at all, more it needs the vrm controller unlocked to suck more current.
Remember I said about my fear over thermal density in particular on air cooling, let's see when voltage is unlocked on the fury-x/ in particular fury.
Looks like the volts are 1.25v (VDroop), thats 50mv more than the 290X
If its not scaling with an overclock the drivers are strangling it, this would explain why it gains so much relative scaling at 4K.
i suspect we ain't going to see a lot of difference between the Fury and Fury-X, certainly not at 1080P.
Soldato
Just tried some memory overclocking.
Stock bench results
1120/550 (crashed at this speed but tried again and it completed)
Keep the core stock and run that again with the memory only mate
nah i hope not
tho i think you have the right for refund if you wanted, esp if they dont have more in stock
its not like uve had it long lol
It is definitely not a driver issue as in unoptimized drivers. At 4K in an intensive game the FuryX still doesn't respond well to overclocks. where the gains should be almost linear.
Amd has done a lot with a power, there are much more aggressive in lower power consumption and adjusting boost clocks. One of the reviews found some very interesting results, as you increase the power tune adjustment the performance actually dropped when overclocking. they found best result with power tun at + 5-10%. That could indicate some kind of bios issue.
The FuryX uses 40W more than the 980Ti, that is an average across many reviews. The FuryX is a very different card to the 290X in terms of power, heat and voltage tolerances. In order to get power within 40W of the 980Ti they had to run the FuryX much cooler with water. The card is designed to run at 60*C vs85-90 of the Hawaii and 980Ti because at those temperatures ti would draw far too much power.
I'm not saying the FuryX wont be able to overclok well, it could be a BIOS issue, but the fact remains at stock it doesn't clock well and what overlocking is possible doesn't return the same kinds of gain one sees in the 980Ti or the 290X.
Yeah I would take the indicated voltage as a rough guide too, if you consider the adaptive vao and avfs which is onboard, also Amd have their Avfs system to setup with 3rd party overclocking/voltage unlocked programs (trixx, msi ab etc)
So this will take time, as I said before too it looks like there's a front end performance issue with the fury at low resolutions, Fury behaves very similar to tonga 285 depending on the game engine.
Whilst performance scaling vs clocks is a factor, there's also clock stability and voltage scaling as the other factor too.
edit vao not vso
Last edited:
Associate
Yup the Hawaii powertune was far more advanced than the Tahiti, it could change the frequency of the card a good few thousand times per sec depending on load...but the Fiji goes one step further...using the feature more aggressively, and switching off not used cores if they not working even it this is just a tiny amount of time
Caporegime
Greg has just posted a 9% performance increase with a 7% core overclock, hows that for scaling? go figure....
Tomb Raider is Direct Compute heavy, not DrawCall.
I just don't think the performance at lower res, or lack there of has anything to do with the architecture, at any res including 1080P Tonga (the same architecture) performs remarkably well, 'relatively' much better than Tahiti and Hawaii it replaced sometimes blowing a way the much bigger Tahiti based 280X.
The bottleneck is not from the architecture, its strangled by AMD's high Driver overheads, DrawCall, its too powerful for their own Drivers, something Windows 10 will fix.
Tomb Raider is Direct Compute heavy, not DrawCall.
I just don't think the performance at lower res, or lack there of has anything to do with the architecture, at any res including 1080P Tonga (the same architecture) performs remarkably well, 'relatively' much better than Tahiti and Hawaii it replaced sometimes blowing a way the much bigger Tahiti based 280X.
The bottleneck is not from the architecture, its strangled by AMD's high Driver overheads, DrawCall, its too powerful for their own Drivers, something Windows 10 will fix.
Last edited:
Associate
Yeah its a shame they had to work hard to push down the power consumption to match the stock Ti/TX, when all the "power consumption is important" 980Ti owners buying the custom cards, then even putting higher TDP bios on them just to see the high clock, and to suddenly forget about the +200W their card is pulling over the stock cards....
Not easy to match the expectations to meet the stock Ti power consumption and beat the power guzzler custom Ti cards on performance
Soldato
Just tried some memory overclocking.
Stock bench results
1120/550 (crashed at this speed but tried again and it completed)
Very nice increase there considering such little overclock.
Be really nice to see what it can do with voltage control.
Caporegime
That is a very good point, TDP is one thing and it means nothing unless its as Nvidia intended and untouched.
Toms Haware put this to the test, if you do interfere with Maxwells Goldilocks reference settings too much the power consumption rockets, their GTX 980 (None TI) overclocked was pulling more power than a Kepler GK110, some factory overclocked GTX 980's pull nearly that much. Some reviews have their Factory overclocked 980's close to a 290X in power consumption.
Last edited:
Caporegime
Power consumption rockets on any card once you start adding voltage and increasing TDP.
Soldato
Yup... I want to this as well, as an engineer said that memory overclock didn't actually work....
Caporegime
Yes but Kepler / Hawaii don't near double their TDP.
My 290 is factory overclocked and overvolted its still well within AMD's set TDP.
Last edited:
Caporegime
Damn it, I just switched back to the TX to record Crysis 3. I will do it later.
I will say though in my half hour or so with the furyx aside from the pump issue the card is totally silent. Can't even hear the fan on the rad even though i had the rad sitting next to me. Build quality is also great, feels very high end and solid, and i'm not one for caring much about how a card looks or feels.
The card also seems to cool down pretty quick, with the 295x2 it remains physically pretty warm to the touch for a while after gaming, this only seems to take a few minutes before its cooled down.
Build quality aside though the pump issue really takes away from that a lot, why go all out to make the card look great and then let QC slip on pumps and some rads that people have got with bent fins? Just seems they took their eye off the ball and concentrated on the card aesthetics while a few other things snuck by unnoticed.
Was meant to receive a sapphire card tomorrow but with this port BS in france its not gonna happen. Just hope when it arrives its also not put back on the next boat back to Germany.
The card also seems to cool down pretty quick, with the 295x2 it remains physically pretty warm to the touch for a while after gaming, this only seems to take a few minutes before its cooled down.
Build quality aside though the pump issue really takes away from that a lot, why go all out to make the card look great and then let QC slip on pumps and some rads that people have got with bent fins? Just seems they took their eye off the ball and concentrated on the card aesthetics while a few other things snuck by unnoticed.
Was meant to receive a sapphire card tomorrow but with this port BS in france its not gonna happen. Just hope when it arrives its also not put back on the next boat back to Germany.