Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,733
- Location
- Planet Earth
I do also think the optimal GCN clockspeeds are sub 1GHZ too - look at the AMD professional GPUs for example.
-
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.
AMD Polaris architecture – GCN 4.0
- Thread starter Boomstick777
- Start date
I do also think the optimal GCN clockspeeds are sub 1GHZ too - look at the AMD professional GPUs for example.
It's going round in circles from both sides in this argument. Fiji was sold as an overclockers dream and it isn't but on the other hand 980Ti does not get 30% actual OC. Andy is taking the lowest boost clocks from 980Ti as a base comparison to max overclocked versions and that is also wrong.
My ACX 2.0 had a max boost at stock of 1304 and drops to ~1217 during normal prolonged gaming sessions. Max OC with voltage and custom fan profile was 1520 but dropped to ~1350 during prolonged gaming sessions. Similarly a reference 980Ti will not hold 1450 for long let alone 1500+
I have seen and read reviews were reference 980Ti was getting 1450+ OC but actual game performance increased only 9% because the reported 25% OC is not actually true.
Last edited:
Agreed and if you look at tonga oc scaling 285/380/x
Fury's clock in the exact same range as Tonga 1100-1150.
The problem with fury is they don't have the front end performance at the low clockspeeds.
At 1000-1050 the power consumption is good but performance is between a 980-980ti.
The fury needs to be in 1250-1300 range to really perform.
As for debating maxwell overclocking ability it's subjective to silicon lottery and how well the.pcb is made. I've seen plenty ref 970's struggle to hit 1450, yet i'vd seen many custom 970's hitting 1500-1550. Mine was flukey and is 1600 game stable.
My friend bought one of thse gibbo.cheap deal imno3d 980ti hybrids, it has a ref pcb he can only get 1480 game stable, yet i've seen 980ti on air scale better.
I was very careful what I wrote above.
Most people consider Kingpin 980 Ti's to be fantastic overclockers. In truth they are crap for overclocking as mine will already boost to around 1450mhz to 1460mhz out of the box. On top of this they will overclock another 120 or 130mhz to reach anything up to 1590mhz on air. The point is they are not great overclockers because they are already factory overclocked.
The Fury Xs on the other hand are not factory overclocked so will overclock better than the Kingpins.
The point I am trying to make here is I mentioned a card (Kingpin) that most people assume without thinking is going to be a great overclocker but in practice has very little left in the tank and compared it to a card (Fury X) which most people assume again without thinking is a poor overclocker but in practice is quite reasonable.
If I compared the Fury X to a reference GTX 980 Ti it would be a different story.
The real problem for the Fury X is it's base performance is poor, not it's overclocking. Having said that it is cheaper than most GTX 980 Ti's so that evens things up.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,222
Still essentially wrong depending on what resolution you base performance on. At stock in most reviews the base performance at 4k/1440p is around the same. So it then comes down to overclocking where the extra 10-20% performance the 980ti can achieve makes it the overall winner. You basically have close to these gains without overclocking but you also would have paid for it.
Anyhow i get what you were trying to say.
If you used the Kingpin for what it is designed for (LN2), you would see a much better clocking card. Running a kingpin on air is fine but if you want 2000Mhz+ clocks, you will need LN2 and this then would be a good overclocking card.
http://www.kitguru.net/components/g...-hits-2-1ghz8-4ghz-with-ln2-sets-new-records/
Edit:
And to make it fair, the Fury X also overclocks much better on LN2 with 1450Mhz and 1000Mhz on the memory
http://www.pcper.com/news/Graphics-Cards/AMD-Radeon-R9-Fury-Unlocked-Fury-X-Overclocked-1-GHz-HBM
Last edited:
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
I don't know what's wrong with your card or case, but neither of my cards throttle, even in SLI
I also never said that everyone would get 30%, I said 20% over reference was pretty much guaranteed and that 1500 was doable by many - your using your factory OC'd cards 1304 as a base line, but that card is already OC'd by 104mhz, its up to you if you want to belittle your own card but if I have a card that's running 1500 compared with the reviewers benchmarking at 1200, I call that a 25% OC over reference, some people get 30% but no where did I say that was the norm
if you can't argue the actual points made, don't go off creating exaggerated strawmen
the lowest boost clock on the reference would be 1075mhz, I'm not using that, I'm using the actual results from actual reviews to compare their actual benchmark results... when I run the same tests at the same settings I get a pretty linear comparison
Last edited:
Exactly this on all counts
Soldato
- Joined
- 5 Sep 2011
- Posts
- 12,881
- Location
- Surrey
One would need to use the card in the right capacity to take advantage of the KPEs level of engineering. Also Maxwell as an architecture does not respond well to voltage and as such most if not all cards will hit a limitation around 1500mhz to 1600mhz or not far short of this. In order to push past this the GPU needs to be ran at sub zero temperatures.
This has even been pointed out by Vince multiple times when speaking about the similarity between one GPU to the next. Lower ambient temperatures just aren't enough, this is through no fault of EVGA or the cards engineering. If the proper research is done these types of things (disappointment) can be avoided.
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
ok, take this as an example - card running at 1500mhz, 36% improvement in actual performance over the reference 980ti
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_Waterforce/26.html
if 36% in actual performance isn't better than a FuryX overclock I will eat my own poo
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_Waterforce/26.html
if 36% in actual performance isn't better than a FuryX overclock I will eat my own poo
Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,733
- Location
- Planet Earth
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
that's not the overclocking results at 1500mhz
try clicking the actual link I posted
102fps for reference 980ti, to 139fps for 1500mhz
I do feel that it is worth mentioning that all the FuryX's are stock reference cards and in my opinion they should be compared to stock reference NVidia cards, with the listed boost clocks taken as the point to calculate from. If a card boosts more than the listed reference clocks, that is just a bonus in my opinion. Also remember the 1050 of the FuryX is the boost clock with powertune doing it job nicely.
I think these days because we have been stuck on 28nm for so long all these cards are pre overclocked.
Back in the early days where a 7950 would be clocked at 850mhz and overclock to 1200mhz, 40%.
Now they are clocked out of the box to 10 or 15% of what they will do max outside of short bench mark runs.
It's a way to increase performance while new tech is not available.
Back in the early days where a 7950 would be clocked at 850mhz and overclock to 1200mhz, 40%.
Now they are clocked out of the box to 10 or 15% of what they will do max outside of short bench mark runs.
It's a way to increase performance while new tech is not available.
Thank you for proving my point, you are taking apples and comparing to oranges to get the best possible numbers. Overclocking performance delta should be compared on the same GPU, ie reference 980Ti stock compared to reference 980Ti OC.
I already accept the Fury X is a pooh overclocker but am arguing on a card by card basis the 980Ti reference is not giving a 25%-30%+ OC. That chart you posted shows an actual in game 11% performance overclock on the GTX980Ti Waterforce compared to stock 980Ti Waterforce. To compare it to a reference 980Ti and saying reference 980Ti is capable of 36% OC is disingenuous.
Fury X £500
Reference 980Ti £530
Gigabyte Waterforce 980Ti £630
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,522
So the fact I can get 1500mhz from my £527 card is irrelevant?
Without any throttling.
I was using the benchmarks at 1200 and 1500 to show the performance delta. What card the reviewer used is not the point, you said that a 980ti running at 25% OC would not show a 25% in FPS
Without any throttling.
I was using the benchmarks at 1200 and 1500 to show the performance delta. What card the reviewer used is not the point, you said that a 980ti running at 25% OC would not show a 25% in FPS
Andy my card is one of the best clocking 970's, it will do over 1600mhz (20%) for about 20 or 30 minutes and then throttle by over 100mhz.
To take 5 minute benchmark runs at highest possible clocks to imply 25% overclocking an average given is peddling the myth.
To take 5 minute benchmark runs at highest possible clocks to imply 25% overclocking an average given is peddling the myth.
Last edited:
I think the point some of the guy trying to explain to you that the cards overclocking potential should compare to the real world performance, not the reference clocks nvidia gave.
If someone hears that maxwells can overcock 25-30% he will assume that adds to the numbers he sees in the reviews. He thinks ref clock result in 100fps, and he will add 30% OC to that. But most reviews doesn't mention the boost clocks of the cards, and as some poited out the cards in reviews usuallu ran at 1300-1350mhz, so when he buys the card and OC it to 1500 he only got around 10% over the FPS numbers seen in the review.
Yes he has 25-30% over the ref clocks, but not compared to real world use.
While on the other hand the Fury runs at 1050mhz in every review, so you if you manage 10% OC you manage it over the performance numbers shown in the review.