I miss the good ol days, pretty sure games were better and **** was more exciting, or perhaps I am just old now
I'd like to see a pic of that
b.) The 7970 scales better with clockspeed increases showing a 1:1 ratio of clockspeed increase to FPS increase 7970 with a 30% overclock gets 30% more performance while the 680 does not a 30% overclock on the 680 nets 18% more performance On a card like the 7970 Lightning built to push the GPU as far as possible, this becomes very relevant.
problem is those are two different sources and testing methods, we can't really draw those conclusions strictly on this little information
Yeah, but it is pretty consistent when you use data just from guru3d as well.
Their 7970 Overclocking Review 1125Core 21% Overclock
Battlefield 3 : 17% improvement with 21% OC
Crysis 2 : 16% improvement with 21% OC
AvP : 16% improvement with 21% OC
Their 680 Overclocking Review 1300Core 29% Overclock
Battlefield 3 : 15% improvement with 29% OC
Crysis 2 : 16% improvement with 29% OC
AvP : 17% improvement with 29% OC
You see the same playing out in OC results elsewhere for the 680.
If you contrast Kepler scaling to Fermi you see that Kepler does not scale nearly as well as Fermi does with overclocking. Another review from guru3d with a GTX 580 overclocked by 28% from 772core to 987core.
There you see Fermi scaling like Tahiti/7970 does with near perfect scaling. 29% improvement in BFBC2 with 28% OC. I'd assume the mem overclock helped here as well with the better than 28% OC to performance result.
Taking Kepler OC results contrasted to the results you'd gain from a good clocking 7970 Lightning would make the 7970 Lightning more appealing to me at the same pricepoint of current 680s, $499. Tahiti scales better with overclocking and the Lightning models are cards built to overclock.
Non-reference 680s may be impressive, but they're going to have to clock really high to give big gains with the less than Fermi-esque scaling Kepler has with core speed increases.
The point still remains that even a reference 7970 will make it to 1200 MHz at least 90% of the time With the stock cooler if you run it loud enough.
That means a 30% avg overclock and an average performance gain of 20-25%. And that is guaranteed 90% of the time.
At $550 it makes no sense but it is perfectly fine at $475-500. And there is no way that a 680 oc will give you noticeable gains in a majority of games than a 25% faster than stock 7970
Also it is IMp to note that if cards get over 75-100 fps the fps become irrelevant, the same is the case if min fps drop below 30-40 fps. Only the remaining cases matter and those are usually at 1080p with 8x aa or 1440p without aa and at those settings almost always both are nearly equal stock to stock and once overclocked the 7970 is a bit faster but not by much. So $500 should be the price for both
Yeah, but it is pretty consistent when you use data just from guru3d as well.
Their 7970 Overclocking Review 1125Core 21% Overclock
Battlefield 3 : 17% improvement with 21% OC
Crysis 2 : 16% improvement with 21% OC
AvP : 16% improvement with 21% OC
Their 680 Overclocking Review 1300Core 29% Overclock
Battlefield 3 : 15% improvement with 29% OC
Crysis 2 : 16% improvement with 29% OC
AvP : 17% improvement with 29% OC
You see the same playing out in OC results elsewhere for the 680.
If you contrast Kepler scaling to Fermi you see that Kepler does not scale nearly as well as Fermi does with overclocking. Another review from guru3d with a GTX 580 overclocked by 28% from 772core to 987core.
There you see Fermi scaling like Tahiti/7970 does with near perfect scaling. 29% improvement in BFBC2 with 28% OC. I'd assume the mem overclock helped here as well with the better than 28% OC to performance result.
Taking Kepler OC results contrasted to the results you'd gain from a good clocking 7970 Lightning would make the 7970 Lightning more appealing to me at the same pricepoint of current 680s, $499. Tahiti scales better with overclocking and the Lightning models are cards built to overclock.
Non-reference 680s may be impressive, but they're going to have to clock really high to give big gains with the less than Fermi-esque scaling Kepler has with core speed increases.
One fatal flaw regarding the whole argument is that the actual overclock is much more like 18-20% rather than 28-29% since the stock GTX 680 is usually boosted to around 1070-1080MHz most of the time, by default.
That's only correct if you are just looking at the offset you define for a 680. You set an offset on the GTX 680 and this offset applies over and above the GPU boost. Guru3D accounted for this and explains it in how he set up the overclock. He set an offset of 200Mhz on top of the 1006stock, with GPU boost this brought the core up to 1300 at peak.
http://www.guru3d.com/article/geforce-gtx-680-overclock-guide/2
The offset may have been 20%, but the core speed increase is regardless of that, 30%. So the point is spot on in that a 30% core increase only brings about 15% more performance. You have to account for GPU boost as well as an overclock to give an accurate indication of clockspeed scaling. In this case it's 1:.5 for clockspeederformance increase.
Kepler just doesn't scale as well as Fermi did with clockspeed. It's about 50% of the improvement in games you saw on Fermi with overclocking, going on guru3D and numbers elsewhere.
Their 680 Overclocking Review 1300Core 29% Overclock
Battlefield 3 : 15% improvement with 29% OC
Crysis 2 : 16% improvement with 29% OC
AvP : 17% improvement with 29% OC
You see the same playing out in OC results elsewhere for the 680.
.
Sorry I am going to point this out to clear the confusion even though this has already been stated earlier in the Kepler Reviews thread. 1300mhz is not a 29% overclock.
A stock card hits up to 1110mhz and sometimes higher. That means a 1300mhz is only a 17% overclock. This is why you see growth of 15-17%. GTX680 does not have poor scaling, it's just that the calculations are made from 1006 mhz stock which is NOT how it functions in games.
It also was investigated that some cards run faster at 1186mhz than at 1250mhz Boost clock because a card with 1186 mhz Boost Clock might boost to 1300mhz+ while a card at 1250mhz may not have any more room for Turbo and perform worse despite what appears to be a higher overclock.
That's why comparing clocks on 680s is trivial. We need to know how far each overclocked 680 can boost up to.
This 1 page exactly explains how GTX680 overclocking works.
Hmm, but I got the idea that GTX 680's maximum boost step at stock is 1110MHz (in that it's not "sometimes higher")? Good post, anyways!!
You negate your own point. Our point was that a 1006 MHz 680 beats a 7970 and the 7970 must pass 1050+ to just equal performance. But according to you a stock goes up to 1100+ and thus it is a 1100 MHz 680 which actually competes with a 1050 MHz 7970. So if both are set at 1300 MHz the 7970 clearly wins
But that is using your logic, IMO if you oc both to 1200-1250 MHz then at 1440p or 1600p they will tie and at 1080p without much aa the 680 oc will be better by an unnoticeable margin.
But based on your logic the 680 has worse ipc not worse scaling
These arguements have run their course. Stock vs. stock the 680 takes a small lead in performance. Overclocked, they are for all intentional purposes the same.
I am saying there are 3 issues at play:
1) GPU temperatures affect GPU Boost
2) Variability in cards themselves
3) Power settings in Precision X.
You can increase the GPU Boost with 0 overclocking by simply raising the power in GPU Boost by 32%. Without even moving the GPU offset, the card will turbo boost to even higher speeds out of the box.
Regardless since GPU Boost varies on a per game, per card, per test bed basis, when we look at a 1300mhz GTX680, it may look like a 30% overclock, but may only be a 17% overclock because some stock 680s easily hit 1110mhz out of the box. Did Guru 3D's card hit 1110mhz or higher in their reference testing in a particular game? We don't know since they never told us. It's inconclusive to say that a 1300mhz GTX680 is 30% overclocked from a reference 680 since it implies the reference card has 0 GPU Boost in games. That's almost impossible unless intentional.
The Boost Clock will vary amongst cards. That's why when some reference cards in reviews hit 1300mhz with Boost, people dismissed those as cherry picking. It's because of variability in GPUs, test bed temperatures, GPU temperatures, etc. that some cards should hit 1300mhz with overclocks while others may never reach beyond 1058mhz.
Since GPU boost declines based on temperatures and you cannot control it unless you lower the temperatures, unless we know the exact Boost frequencies a stock card achieved, looking at % overclock is inaccurate.