Shootout: 780 Lightning vs 290

Elfear

Diamond Member
May 30, 2004
7,097
644
126
I bought a 780 Lightning awhile ago because I wanted to try something new and prices had dropped a lot on the high-end 780's. It turned out to be a great card and I would have kept it longer but the mining craze hit and I couldn't resist. I ended up buying a pair of reference Powercolor 290s to replace it.

I decided to do a shootout between the Lightning and one of my 290s since it's been a hot topic on the forum the last couple months. I don't have all the modern games in my library but I do have a few and some classics that still give modern GPUs a workout.

Benchmark Rig

OS: Win 8.1 64-bit
Drivers: AMD - 13.11 Beta 9.5, Nvidia - 331.65
CPU: 4770k@4.7Ghz 1.345V HT off
Game Drive: Samsung EVO SSD

Testing Methodology

Each game was run 2-3 times at each listed setting and the best run was used in the charts below. There was very little variability between runs so I'm confident in the results.

Average watts were taken with my calibrated eyeball and a Kill-O-Watt during the last run to make sure both cards were warmed up. Power consumption fluctuated around but generally it was within 15-20W of the average.

For most of the testing I used the stock cooler for both cards. For max oc and a couple other benches I used an EK full water block on the 290. You'll notice on some of the charts that I have "290@1150/1450 H2O" listed to compare power consumption between the reference cooler and the watercooled card. Taking the blower fan out of the equation obviously reduces power draw but the lower temps also help a lot too.

I decided to bench at three different overclocked settings:
Stock - Out-of-the-box performance
Gaming - Clocks and voltage I'd be comfortable with 24/7
Max OC - Bench only settings.

In general the clocks for both cards were the following:

780

Stock - 1124/1500
Gaming - 1340/1650 (+100mV or 1.2V after droop)
Max - 1430/1675 (+200mV or 1.29V after droop)

290

Stock - 975/1250
Gaming - 1150/1450 (+60mV or 1.18V after droop)
Max - 1250-1285/1400 (+219mV or 1.29V after droop)


General Observations

290

My computer is in an open case and sits about three feet away from me. From my limited time using the reference 290 cooler I will say that at 47% or below the fan wasn't that bad. About the same as the reference AMD coolers I've used in the past. Once I started ramping up fan speed though it got very loud. At 90% I didn't want to be in the same room.

The reference PCB is fairly weak IMO with it's 6-phase design. Having used a few reference 7970's in the past I was disappointed with the capability of the 290 PCB. The "leakier" card (80.5% ASIC) I could put ~1.39V through (1.29V with vdroop) and the less leaky card (72% ASIC) could only take 1.34V (1.25V after droop) before the image got all fuzzy and started crawling. I've never seen that before when overclocking too high, usually it either artifacts when the core is tapped out or goes checkered or freezes if the memory is too high. All I can think is that the VRMs just couldn't handle the voltage and it corrupted the video output (only happened on the dual-link DVI).

Even under water with a block known for it's VRM cooling (EK) the VRM's still hit 70C at 1285/1400 1.29V looping Heaven. Needs more phases!

780

This card is an overclocker's dream. 20-phase power delivery (vs. 8-phase on the regular 780s) and comes with 2x 8-pin power connectors. The cooler is capable of dissipating 550W of energy and is still quiet to boot. While benching at 1420-1440Mhz 1.35V (1.29V after droop) the core only hit 78C and VRMs only reached 62C (fans at 80%/70%/80%).

Recommendation

When I first started the shootout it wasn't really a fair comparison because of the large price difference ($380 vs $515 AR). The deck was stacked against the 290 with its reference PCB and stock cooler compared to the Lightning's tank-like build quality and very efficient cooler. I was willing to cut the 290 some slack if it didn't turn out quite as good as the 780 because of the price.

With today's prices however, it's a different story. The 780 Lightning is going for $500 AR and a reference 290 is now the same price with all the gouging going on.

As you'll see in the charts below the cards perform very close to each other, mainly due to the excellent overclocking ability of the Lightning. At today's prices I would recommend the 780 Lightning because it's just such a well-built card. In fact I don't think I've ever used such a quality built video card and I must say it makes going back to reference cards tough. If prices drop down to MSRP or better, I would have a harder time recommending the 780 but prices are what they are so the nod goes to the Lightning.
 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
Metro 2033
At stock the 290 throttled down to 666Mhz w/o DOF and 763Mhz w/ DOF a couple times for ½ a second.








 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
Crysis Warhead
The 290 throttled while benching to:
Stock – 901Mhz a couple times
1150/1450 – 1088Mhz regularly
1265/1450 – 1231Mhz a few times




Avalanche









Frost





Crysis

Custom cvar (sorry can't remember all the settings but it is basically maxed).



 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
Far Cry 2

This game can be run very easily on a low- to mid-range GPU these days so this bench was just because I had the game and it has a built-in benchmark.








Skyrim



Mods






Borderlands 2





 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
Crysis 3
Settings - All maxed, 2xSMAA, no motion blur
290 throttled down to 1082Mhz occasionally during the 1150/1450 run.









Battlefield 3









 

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
First I want to say grats on 4.7GHz ;)

Secondly great job on the comparison!

Lastly I want to ask you how much you think the memory was holding back the 780 at 1600p?

I ask because in BioShock I was able to beat Face2Face'es score with lower clocked core with my card, I wasn't sure if that was due to the fact that my vram went higher, or because my Haswell was at 4.8GHz and his Ivy was at 4.8GHz.
 

24601

Golden Member
Jun 10, 2007
1,683
39
86
Amazingly detailed review!

Was "AMD Optimized Tessellation" and "Surface Format Optimization" turned off for these tests?

Was Texture quality set in drivers to "High Quality" for both drivers?

Was "Trilinear Filtering Optimization" turned off for both drivers?

Thanks!
 
Feb 19, 2009
10,457
10
76
Nice work on the test. I've come to same conclusion the R290/X reference cooler AND the 6 phase power is really awful for big OC, even on water they rarely get above 1.25ghz.

I would have waited for custom cards, but the mining profitability didn't wait for me, so I grabbed these reference R290s, something I resisted in the past, to never buy reference AMD cards ever due to their horrible cooler. But now both R290s are paid for and I have a big pile of coins left over, so in that sense, it was an awesome buy. I could ebay these now (without losing $!) and grab custom R290s when its out here next month, at MSRP to boot. :)
 
Last edited:

Teizo

Golden Member
Oct 28, 2010
1,271
31
91
A little birdie told me you in the process of doing this. Good right up man. Thanks for sharing.
 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
First I want to say grats on 4.7GHz ;)

Thanks Balla. I picked up the 4770k when MC has their big sale last month. Not as good as your golden chip but better than the 4670k it replaced.

Secondly great job on the comparison!

Lastly I want to ask you how much you think the memory was holding back the 780 at 1600p?

I ask because in BioShock I was able to beat Face2Face'es score with lower clocked core with my card, I wasn't sure if that was due to the fact that my vram went higher, or because my Haswell was at 4.8GHz and his Ivy was at 4.8GHz.

I don't think the memory bandwidth was holding it back too much in Bioshock. If I still had the card, I could bench it at different memory speeds and see how it affected performance. Face2Face just picked up a 1440p monitor so maybe he could test for you.


Amazingly detailed review!

Was "AMD Optimized Tessellation" and "Surface Format Optimization" turned off for these tests?

Was Texture quality set in drivers to "High Quality" for both drivers?

Was "Trilinear Filtering Optimization" turned off for both drivers?

Thanks!

Thanks!

I left all driver settings in their out-of-the-box configuration. From what I've read in different reviews that generally gives comparable results between the red and green teams.

Nice work on the test. I've come to same conclusion the R290/X reference cooler AND the 6 phase power is really awful for big OC, even on water they rarely get above 1.25ghz.

I would have waited for custom cards, but the mining profitability didn't wait for me, so I grabbed these reference R290s, something I resisted in the past, to never buy reference AMD cards ever due to their horrible cooler. But now both R290s are paid for and I have a big pile of coins left over, so in that sense, it was an awesome buy. I could ebay these now (without losing $!) and grab custom R290s when its out here next month, at MSRP to boot. :)

Went through the exact same conundrum. I was set to wait for custom cards but mining sucked me in. Can't wait for the 290X Lightning to surface.
 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
A little birdie told me you in the process of doing this. Good right up man. Thanks for sharing.

Thanks. I would have had it out a week earlier but trying to find the sweet spot on the 290 took awhile. It really needs a better PCB for uber overclocking.
 
Feb 19, 2009
10,457
10
76
Thanks. I would have had it out a week earlier but trying to find the sweet spot on the 290 took awhile. It really needs a better PCB for uber overclocking.

Its a real shame that the Sapphire Triple cooler is using the same reference PCB with 6 phases, a massive letdown since its got heaps of thermal room for OC.

The ASUS model have custom mosfets and 8 phases so thats a bit extra. But really, gotta get beefier models to push OC higher. Can't really muster up versus Lightning's 20 phases!
 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
Its a real shame that the Sapphire Triple cooler is using the same reference PCB with 6 phases, a massive letdown since its got heaps of thermal room for OC.

The ASUS model have custom mosfets and 8 phases so thats a bit extra. But really, gotta get beefier models to push OC higher. Can't really muster up versus Lightning's 20 phases!

Ya the high-end 290Xs should stretch their legs pretty well with beefier PCBs. Overclocking is really a much better experience with a robust power delivery system. The Lightning was rock solid until you hit the limits of the core or memory and then you backed it off and benched. No drama. With less power phases you get instability much more quickly and it doesn't seem to have any rhyme or reason. My 6-phase 7970's were easier to clock so it must be the power requirements of Hawaii.
 

Jaydip

Diamond Member
Mar 29, 2010
3,691
21
81
That was an awesome job dude well done, I think the verdict was known to all of us though an oced 780 is a monster :D
 

DarkKnightDude

Senior member
Mar 10, 2011
981
44
91
Very impressive showing by both cards. 290 is obviously the best bang for buck here. 780 is a good OCer.
 

Majcric

Golden Member
May 3, 2011
1,369
37
91
I'm in the monority of folks who usually buy reference cards. However, both the GTX 780 and the 290 are exceptions. If i'm buying one these two cards it is going to aftermarket for sure. For the 780 a classy or lightning, the 290 it doesn't matter.
 

DarkKnightDude

Senior member
Mar 10, 2011
981
44
91
Yes you're right, he's in Taiwan!

Good morning!

Late night here. :biggrin:

Yeah, prices here are more based around MSRP then say the US usually. Though there has been price rises here lately noticeably, I blame the local newspapers for going on about the miner craze which definitely has not helped.

Nvidia cards here are priced more then their US counterparts though, price-performance wise.

Anyways, offtopic.