Shootout: 780 Lightning vs 290

[blackened23]

You can unlock the voltage on your card up to 1.3V, fastX, with this afterburner trick:

http://www.overclock.net/t/1398725/...its-on-lots-of-cards-titan-to-gtx460-with-llc

You can use the SkyNET BIOS as well but I would not recommend it. A lot of folks are mentioning that their cards do not properly go into "idle" modes at the right times with the skynet BIOS, which would of course lead to high idle temps.

 

[BallaTheFeared]

Using Skynet and the 1.3v unlock with my card, no idle issues for me.

Edit: Using Auto for desktop, if you manually set the voltage it sticks even at desktop. Auto vs 1.2v is about 10w difference at idle.

Use MSI's 2D/3D profile to automatically switch between forced/0.9v

Also got my old screen up to 74 Hz with the Nvidia tool

G-Sync will be my next side grade.

 

[Fastx]

You can unlock the voltage on your card up to 1.3V, fastX, with this afterburner trick:

http://www.overclock.net/t/1398725/...its-on-lots-of-cards-titan-to-gtx460-with-llc

You can use the SkyNET BIOS as well but I would not recommend it. A lot of folks are mentioning that their cards do not properly go into "idle" modes at the right times with the skynet BIOS, which would of course lead to high idle temps.

Blackend thanks for that! I just skimmed though it now but will take a closer later. If this does what it looks like this would be great and something I was looking for but had no luck yet. That is to be able to raise the voltage (higher offset) just little higher if needed for over 1200+ with out having to do a BIOS change. I was reading about classified voltage controller earlier real quick but believe I need a BIOS change for that. I'm really not interested in a BIOS change least at this point because the SC ACX is single BIOS card. After I ordered my card I read in a forum FTW was a dual BIOS had I know this this before I most likely would have went with the FTW card.

I bought this card with the 967/1020 and around 1120 in game boost per HC so that is decent for me to start off from looking at charts, but around 1210-1240 would be nice icing on the cake for me imo. Then I won't be looking back at all if you know what I mean and I am not talking about my CFX. Good to see Termie's 780 SC ACX was able to also clock over 1200 gives more added possible hope.

 

[Elfear]

That said, this doesn't change the fact that the vast majority of reviews published for overclocking on the 290X/290 vs GK110 heavily favor the latter for maximum overclocks, overclock scaling, and clockspeeds above stock.

GK110 does reach a higher max overclock vs Tahiti and Hawaii in general. Scaling though seems to be pretty close and even favors Hawaii. I threw together a chart from the benchmarks I ran (stock vs. gaming clocks):



Total scaling across all games was 82% for the 290 and 76% for the 780. I'm assuming the >100% scaling for BL2 and Warhead Avalanche were because of benchmark variability and/or minor throttling at stock with the 290.

It is 400Mhz above stock. The HOF 780 achieved speeds in excess of 1300mhz at HardOCP. The lightning GTX 780 achieved 1333MHz. 433MHz above stock. The stock boost of the reference GTX 780 is 900MHz.

GTX 780 specifications:

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-780/specifications


GTX 780 GPU Engine Specs:
2304CUDA Cores
863Base Clock (MHz)
900Boost Clock (MHz)
160.5Texture Fill Rate (billion/sec)
GTX 780 Memory Specs:

You're kind of comparing apples and oranges though. Even when clocks are pretty close to the same, the 780 still has to overcome a 10% performance deficit. In the same chart you can see a 290@1000Mhz is 18% faster than a 780 at stock boost.

So the 780 does have a lot of headroom if you want to start at 900Mhz but it is also much slower.

 

[BallaTheFeared]

I seem to remember them having 780 and "uber" 780, I believe the normal 780 ran sub 900MHz, perhaps 880?

 

[blackened23]

G
You're kind of comparing apples and oranges though. Even when clocks are pretty close to the same, the 780 still has to overcome a 10% performance deficit. In the same chart you can see a 290@1000Mhz is 18% faster than a 780 at stock boost.

So the 780 does have a lot of headroom if you want to start at 900Mhz but it is also much slower.

Say what?. The stock 290 in reference form is basically tied with the 780. Once you overclock, various aftermarket 780s are hitting speeds 15-20% faster than Titan. There is no 10% performance deficit when comparing a stock 780 to a stock 290. You are comparing a factory OC'ed 290 to a reference 780. Uh. We could compare factory OC'ed 780s which have already hit speeds in excess of 15-20% faster than Titan. Which, since the 290X is even with the Titan, makes the factory OC'ed 780 15-20% faster than the stock reference 290X.

This GTX 780 is 20% faster than stock Titan and stock 290X, 30% faster than stock GTX 780:

http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_780_Direct_Cu_II_OC/29.html

The 290 is not hitting 15-20% faster than Titan. Yet this aftermarket 780 starts out 10% shy of Titan. Once overclocked, is 20% faster than Titan. So the 780 aftermarket cards are gaining 25-30% in performance once overclocked.

The gap widens even further once you look at factory OCed 780ti cards:

http://www.hardwarecanucks.com/foru...gigabyte-gtx-780-ti-ghz-edition-review-8.html

This 780ti is 29.5% faster than stock Titan speeds, 29.5% faster than stock 290X uber mode speeds.



IMO that is superior overclock scaling as compared to the 290/X which is struggling to hit overclocks at around 1125-1175 from every review i've seen.

Someone show me the 290 or 290X hitting 29% faster than Titan, or 5-7% faster than GTX 690 speeds. The windforce 780ti is. It is even faster than the 690. Or we could just look at the plain jane GTX 780 - The 780 DC II is hitting 20% faster than Titan. My point here is, the 290 is a great card in aftermarket varieties. I'm not saying otherwise. If these aftermarket 290 cards are selling at MSRP which could be questionable in the states, but if they do sell at MSRP. They will be fantastic cards. However in terms of OC headroom? It is worse than the GK110 by a considerable margin. Maybe a future custom 290X can change that, but the DC II 290X hasn't - most of those cards hit a wall between 1100-1200 from what i've seen. Yet more custom cards would be welcomed and could certainly change things - but as far as I can tell from published data, the GK110 has more tinkering/overclock potential at the moment.

 

[Mondozei]

Blackened, we are not discussing 780 Ti so mixing it in makes no sense.

You are right about your main point. GTX 780 is a beast of card for overclocking with the right cooler. Hawaii is more GPU-constrained even if there is more than ample thermal headroom left.

The 290 is still a better buy. Why? In two words: the future.
We can already see it in VRAM intensive games like Metro:LL. If you look at nextgen games they will all require more and more VRAM. The new Tom Clancy game "The Division" is a good example of that.

Also, the console wins will matter a lot, just look at BF4. Sure you have G-sync but you'd want a 4K 60 hz monitor for that anyway since 4K gaming will be mainstream within 2 years at most and why buy another monitor. So g-sync is 2-3 years out anyway.

 

[Imouto]

Oh mai. Again comparing reference to custom when it suits.

Custom 290X OC -> Custom 780Ti OC = +10%

 

[amenx]

The 290 is still a better buy. Why? In two words: the future.

Future doesnt mean much for those who change cards every new card gen.

We can already see it in VRAM intensive games like Metro:LL.

Metro LL is not a vram intensive game, is just favors AMD cards in this instance.

http://www.guru3d.com/articles_pages/metro_last_light_graphics_performance_review_benchmark,8.html

 

[Gloomy]

Linus tech tips did a reference 290 under water versus GTX 780 reference under water:

https://www.youtube.com/watch?v=RqaHh-y51us&feature=*********&a

GTX 780 reference: 4 phase PCB, GPU boost 2.0, 1.2V voltage limit
R9-290 reference: 6 phase PCB with AMD's boost, 1.25V voltage limit (?")

Net gain: GTX 780 gained 9% at maximum overclocks
Net gain: R9 -290 gained 7% at maximum water overclocks

GTX 780 @ 35C plus/minus a few at 100% GPU load. R9-290 @ ~45C+/- a few degrees @ 100% GPU load.. Gear used:

Koolance Gear
R9 290 Block: Koolance
GTX 780 Block: Koolance
Quick Disconnect Fittings: Koolance
Tube Reservoirs: Koolance
DDC Pump Top: Koolance

Maximum overclock with 24/7 stability in all games - core clock actual: GTX 780 ~1180MHz, R9-290 1110MHz under water per clock listing master spreadsheet:

https://docs.google.com/spreadsheet...by1jNTFvaTF2UHhaMzdMQ0FNM2c&usp=sharing#gid=0

Uhh... he didn't give the 290 any voltage, but he gave the 780 38mV. What gives?

 

[Leadbox]

Say what?. The stock 290 in reference form is basically tied with the 780. Once you overclock, various aftermarket 780s are hitting speeds 15-20% faster than Titan. There is no 10% performance deficit when comparing a stock 780 to a stock 290. You are comparing a factory OC'ed 290 to a reference 780. Uh.

No he's not, please look again
If you're going to dismiss it as a tie, then you should do the same for aftermarket OC Tis' vs aftermarket 290X OC :awe:

 

[el etro]

Say what?. The stock 290 in reference form is basically tied with the 780.

No way. Only in Alice's World(Techpowerup) and at 1080p only. Same goes to Titan vs 290x.
Hardware Canucks point 290 above gtx 780 and 290xDCU tied with 780Ti(But 290x can't catch 780Ti at overclocks).

 

[parvadomus]

Say what?. The stock 290 in reference form is basically tied with the 780. Once you overclock, various aftermarket 780s are hitting speeds 15-20% faster than Titan. There is no 10% performance deficit when comparing a stock 780 to a stock 290. You are comparing a factory OC'ed 290 to a reference 780. Uh. We could compare factory OC'ed 780s which have already hit speeds in excess of 15-20% faster than Titan. Which, since the 290X is even with the Titan, makes the factory OC'ed 780 15-20% faster than the stock reference 290X.

This GTX 780 is 20% faster than stock Titan and stock 290X, 30% faster than stock GTX 780:

http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_780_Direct_Cu_II_OC/29.html

The 290 is not hitting 15-20% faster than Titan. Yet this aftermarket 780 starts out 10% shy of Titan. Once overclocked, is 20% faster than Titan. So the 780 aftermarket cards are gaining 25-30% in performance once overclocked.

The gap widens even further once you look at factory OCed 780ti cards:

http://www.hardwarecanucks.com/foru...gigabyte-gtx-780-ti-ghz-edition-review-8.html

This 780ti is 29.5% faster than stock Titan speeds, 29.5% faster than stock 290X uber mode speeds.



IMO that is superior overclock scaling as compared to the 290/X which is struggling to hit overclocks at around 1125-1175 from every review i've seen.

Someone show me the 290 or 290X hitting 29% faster than Titan, or 5-7% faster than GTX 690 speeds. The windforce 780ti is. It is even faster than the 690. Or we could just look at the plain jane GTX 780 - The 780 DC II is hitting 20% faster than Titan. My point here is, the 290 is a great card in aftermarket varieties. I'm not saying otherwise. If these aftermarket 290 cards are selling at MSRP which could be questionable in the states, but if they do sell at MSRP. They will be fantastic cards. However in terms of OC headroom? It is worse than the GK110 by a considerable margin. Maybe a future custom 290X can change that, but the DC II 290X hasn't - most of those cards hit a wall between 1100-1200 from what i've seen. Yet more custom cards would be welcomed and could certainly change things - but as far as I can tell from published data, the GK110 has more tinkering/overclock potential at the moment.

http://www.anandtech.com/show/7601/sapphire-radeon-r9-290-review-our-first-custom-cooled-290/5

OC performance delta vs stock 780:
Metro LL: 29% faster
COH 2: 52% faster
Bioshock Infinite: 11,7% faster
BF3: 5,6% faster
Crysis 3: 15,5 % faster
TW Rome 2: 23,7% faster

There you have it, an OCed 290 is faster than a 780Ti in games like Metro LL and Coh 2 and probably faster than GTX690 as well. The gtx780 seems to be in the same bag as 290, and 290X is between Titan and 780Ti. 780 seems to have less horsepower per clock than 290, but its tesselation performance and higher OC headroom helps it, however it cant make a difference in the average game scenario. 290X is like 5% slower than 780Ti at the same clockspeeds.
But keep throwing your cherrypicked benchmarks about BF3 and Crysis 3 or whatever. :|

 

[Imouto]

No way. Only in Alice's World(Techpowerup) and at 1080p only. Same goes to Titan vs 290x.
Hardware Canucks point 290 above gtx 780 and 290xDCU tied with 780Ti(But 290x can't catch 780Ti at overclocks).

The fun thing is that even a GTX 780 Classy at 1.32 Ghz is 22% faster than the reference one at stock. If we go to the latest Sapphire 290 Tri-X review by Anandtech the performance versus a ref 290 in Crysis 3 is 17% better while overclocked.

Just add a 22% in those Crysis 3 numbers and we will have a nice score of 63.7 FPS for a OCed GTX 780. An OCed GTX 780 is still 6% faster than a 290 which is way, way off than the stuff some people is trying to make up.

If we go by computerbase's numbers in a 5 games mix for the 290 Tri-X the stock peformance for this card is 18% better than a stock GTX 780 while still having room for a ~12% improvement by Anandtech's numbers.

 

[el etro]

The fun thing is that even a GTX 780 Classy at 1.32 Ghz is 22% faster than the reference one at stock. If we go to the latest Sapphire 290 Tri-X review by Anandtech the performance versus a ref 290 in Crysis 3 is 17% better while overclocked.

Just add a 22% in those Crysis 3 numbers and we will have a nice score of 63.7 FPS for a OCed GTX 780. An OCed GTX 780 is still 6% faster than a 290 which is way, way off than the stuff some people is trying to make up.

If we go by computerbase's numbers in a 5 games mix for the 290 Tri-X the stock peformance for this card is 18% better than a stock GTX 780 while still having room for a ~12% improvement by Anandtech's numbers.

My critic goes to Blackened insisting in "state" that GTX 780 and Titan have equal performance than their AMD counterparts. Thats not true. In case of 290 vs 780, even the OP confirms it. Performance per clock of R9 290 is between Titan and 780, and 290x performance per clock is closer to 780Ti's than Titan's(That's why 290xDCU can outperform the reference 780Ti at 1600p).



Yes, at overclocks 780 goes much more far from stock than R9 290 Goes, but Nvidias card is favored by a bigger golden rate and by a Top-Class PCB(This is a thing that no custom R9 290 have at this moment, knowing the first custom cards are most of times based on original PCB). And the deal of ones that not end with golden samples is to end with clocks like these(That is good clocks, but not the 1300-1400Mhz that many ones state here).

 

[blastingcap]

If 780 3GB and 290 4GB cards were equal price, which they are not, I would choose the R9 290 every time. Why? The 290 has 4GB VRAM, and the 780 has 3GB VRAM. PS4 has 8GB unified memory, and only part of that is needed for OS and game code, which means that a LOT of that memory can be dedicated to graphics. I think we may soon see console ports that require more than 3GB VRAM to run well.

Let's face it, the performance differences between a R9 290 and 780 aren't that big, but the performance difference between having enough VRAM and not having enough VRAM is HUGE.

 

[Fastx]

If 780 3GB and 290 4GB cards were equal price, which they are not, I would choose the R9 290 every time. Why? The 290 has 4GB VRAM, and the 780 has 3GB VRAM. PS4 has 8GB unified memory, and only part of that is needed for OS and game code, which means that a LOT of that memory can be dedicated to graphics. I think we may soon see console ports that require more than 3GB VRAM to run well.

Let's face it, the performance differences between a R9 290 and 780 aren't that big, but the performance difference between having enough VRAM and not having enough VRAM is HUGE.

This has crossed my mind, so how long do you think it will be when 3GB will not be enough and the need to upgrade to a card with more VRAM will become important enough to want to upgrade?

Also at that time do you think 4Gb will be enough?

 

[Gloomy]

within two years I think is reasonable, so a moot point imo (just upgrade at that time)

 

[ICDP]

The extra 1GB is nice to have but I believe it will only be an issue for those who go multi GPU. 1X 3GB GPU will hit a GPU limit long before a VRAM limit becomes a reality. 2X top end GPUs on the other will be sufficient GPU power for around 3 years IMHO, but if a VRAM wall surfaces they become useless instantly.

Look at GTX 580 1.5GB cards. In SLI (or single 590) they are about as fast (or faster) on paper than a GTX780 but in reality the 1.5GB VRAM cause them to hit a massive VRAM wall that cannot be overcome without seriously dropping the settings.

http://www.anandtech.com/bench/product/975?vs=827

If going multi GPU is part of your future plans then IMHO it would be wise to get as much VRAM as affordable to future proof. It may never be a problem but that is what future proofing is all about. Making sure it doesn't.

 

[Fastx]

The extra 1GB is nice to have but I believe it will only be an issue for those who go multi GPU. 1X 3GB GPU will hit a GPU limit long before a VRAM limit becomes a reality. 2X top end GPUs on the other will be sufficient GPU power for around 3 years IMHO, but if a VRAM wall surfaces they become useless instantly.

Look at GTX 580 1.5GB cards. In SLI (or single 590) they are about as fast (or faster) on paper than a GTX780 but in reality the 1.5GB VRAM cause them to hit a massive VRAM wall that cannot be overcome without seriously dropping the settings.

http://www.anandtech.com/bench/product/975?vs=827

If going multi GPU is part of your future plans then IMHO it would be wise to get as much VRAM as affordable to future proof. It may never be a problem but that is what future proofing is all about. Making sure it doesn't.

Trust me when I say this exact thing above (SLI) has also cross my mind.

 

[blackened23]

Uhh... he didn't give the 290 any voltage, but he gave the 780 38mV. What gives?

Correct me if i'm wrong, because I haven't used the new powertune CCC yet. But from what i'm reading, he set powertune to 25% which is 5% beyond the typical 20%. Basically, from my understanding, this ups the maximum dynamic voltage that powertune typically allows at stock settings. . If you adjust powertune beyond 20%, the dynamic voltage at maximum capability goes up. Afterburner does the same thing, but in a more straightforward fashion. Afterburner adjusts powertune in the background, but is more straightforward in the sense that you input "XX" voltage. It ups the "allowable" dynamic voltage allowed by the Hawaii chip, thus allowing for more OC headroom.

I personally thought this was over-voltage beyond stock. Again...Linus used +25% powertune. Because I've seen this mentioned in many 290/X reviews. Now if you've used prior AMD cards, powertune did not do the same thing - the "old" powertune simply capped at 20% and didn't adjust voltage. But the new powertune also adjusts Hawaii's dynamic voltage. And if you go past 20%, it allows voltage above the maximum dynamically. At least this was my impression. Seen here:

This was my impression of the new powertune based on everything I read from 290X launch reviews. So when I saw linus put 25% powertune , he upped the dynamic voltage beyond maximum. Now here's the thing. If you up powertune (from my rough understanding), it will also increase the maximum allowable dynamic voltage at 100% GPU load - but it doesn't spell out EXACTLY what voltage that is. Whereas, afterburner more or less tells you "Hey! you're upping dynamic voltage by XX amount". At least that's my impression. Do correct me if i'm wrong, i'm basing this (powertune functionality for Hawaii) off several articles that i've seen. So i'm merely asking. This is definitely different than the powertune used with the old vanilla 7970, which I was familiar with - that one capped at 20% and didn't have anything to do with dynamic voltage.

If you could clarify this, that would be great. Because i've read this at more than a couple of websites.

 

[Gloomy]

Correct me if i'm wrong, because I haven't used the new powertune CCC yet. But from what i'm reading, he set powertune to 25% which is 5% beyond the typical 20%. Basically, from my understanding, this ups the maximum dynamic voltage that powertune typically allows at stock settings. . If you adjust powertune beyond 20%, the dynamic voltage at maximum capability goes up. Afterburner does the same thing, but in a more straightforward fashion. Afterburner adjusts powertune in the background, but is more straightforward in the sense that you input "XX" voltage. It ups the "allowable" dynamic voltage allowed by the Hawaii chip, thus allowing for more OC headroom.

I personally thought this was over-voltage beyond stock. Again...Linus used +25% powertune. Because I've seen this mentioned in many 290/X reviews. Now if you've used prior AMD cards, powertune did not do the same thing - the "old" powertune simply capped at 20% and didn't adjust voltage. But the new powertune also adjusts Hawaii's dynamic voltage. And if you go past 20%, it allows voltage above the maximum dynamically. At least this was my impression. Seen here:

This was my impression of the new powertune based on everything I read from 290X launch reviews. So when I saw linus put 25% powertune , he upped the dynamic voltage beyond maximum. Now here's the thing. If you up powertune (from my rough understanding), it will also increase the maximum allowable dynamic voltage at 100% GPU load - but it doesn't spell out EXACTLY what voltage that is. Whereas, afterburner more or less tells you "Hey! you're upping dynamic voltage by XX amount". At least that's my impression. Do correct me if i'm wrong, i'm basing this (powertune functionality for Hawaii) off several articles that i've seen. So i'm merely asking. This is definitely different than the powertune used with the old vanilla 7970, which I was familiar with - that one capped at 20% and didn't have anything to do with dynamic voltage.

If you could clarify this, that would be great. Because i've read this at more than a couple of websites.

Powertune governs the maximum power the card can use, not the voltage. And the max is 150%, lol.

 

[blackened23]

Anyway, the maximum powertune was 20% on the 7970 vanilla back in 2012 - I used those cards and of course overclocked them quite a bit. So I wasn't sure what the maximum was. You're certain it doesn't adjust dynamic voltage? I'm only asking. I'm not 100% sure on how the new powertune works.

If you're stating it only adjusts maximum power, i'll take that answer at face value. Then again, that's rather worthless outside of throttle prevention in power virus type of programs such as furmark.

 

[Imouto]

Anyway, the maximum powertune was 20% on the 7970 vanilla back in 2012 - I used those cards and of course overclocked them quite a bit. So I wasn't sure what the maximum was. You're certain it doesn't adjust dynamic voltage? I'm only asking. I'm not 100% sure on how the new powertune works.

If you're stating it only adjusts maximum power, i'll take that answer at face value. Then again, that's rather worthless outside of throttle prevention in power virus type of programs such as furmark.

Every time I've read about Powertune related to Hawaii its duty is to lower voltages and frequencies, not the other way around.

But would be great if you provide any single proof on these cards overvolting beyond shipped defaults instead of making absurd claims and asking other users to prove you wrong.

 

[BallaTheFeared]

There seems to be some confusion about 780 scaling, hopefully despite being at 1080p I can clear that up a bit.

900/1500 @ 1v 260w peak

1100/1700 @ 1.1v 330w peak

23% core OC netted 21% increase in performance. This was with Rome 2, which was necking on me despite not going higher with the OC. An OC which is actually lower than what this 780 GHz boosts to out of the box (but still has stock vram).

Nvidia doesn't just pick a random memory frequency, the cards are fairly well balanced. Increasing core without increasing memory will yield poorer scaling than increasing both.