R9 290 vs GTX 780 - discussion thread

[TechFan@ic]

This doesn't make sense, why was the GTX 780 over-volted while the R9 290 wasn't ?
AMD's overclocking advantage lies entirely in the fact that their cards can be over-volted significantly more than the Kepler based Nvidia GPUs due to Nvidia's greenlight program.

[Gunjob]

This doesn't make sense, why was the GTX 780 over-volted while the R9 290 wasn't ?

AMD's overclocking advantage lies entirely in the fact that their cards can be over-volted significantly more than the Kepler based Nvidia GPUs due to Nvidia's greenlight program.

There's a link to my R290x OC'ing in my sig', I only needed 65mV to achieve a 20% OC. I'm alittle unsure of @LinusTech OC on the 290, and while underwater I at the very least expected some overvolting, I'm aware there is abit of the luck of the draw about overclocking but when all his benches are with OC'd GPU's I'd like abit of talk about each overclock when doing these videos since I feel thats fairly important when doing these sorts of benchmarks. Felt more like a Koolance ad roll than an actual benching video. 

[VSG]

Alright it's finally here. The stunning conclusion of our R9 290 vs GTX 780 showdown. We used a couple of our most taxing games to validate our overclocks, so if they run in these ones, they'll be good to go in others. We hit a hard wall with both cards where if we increase voltage, clock speed, or power limit any further we were not stable, so this is about the best we can do, but as always, YMMV. Here's the link to our latest GPU clock speed spreadsheet. I'm going to email Luke right now to find out why the water cooled clocks aren't on it yet... But they'll be there soon if they aren't already

 

 

Kinda feel you guys did not do this part really, power limit was set to just +25 (R9 290)/+6 (GTX 780), no overvolting done on the 290. How are clock speeds deemed stable- game stable or benchmark stable? Hopefully the former since that's more real-life. 

 

The issue is that I have now seen about a 100 guys comment saying "Ok, so I will now grab a GTX 780 and waterclock it for best performance" or something like that even though you clearly state your particular cards do not in any way guarantee how other cards will perform. So, in order to be fair, the same overclocking parameters should have been used on both cards- either overvolt both to the same degree (and explain why one can go further) and set the power limits the same or do not overvolt at all for both cards.

 

Lastly: You report what I assume was average core temperature for both cards. What about the VRM temps? They play a very big role in overclocking and not all water blocks are created equal here. Koolance, for example, had quickly issued a revision 2 of their water block for the R9-290(x) cards which is a shorter block and apparently deals better with VRMs.

[TheMissxu]

780 still has PhysX, Better 3D, CUDA, Shadowplay, and what does AMD Have ? Not even Mantel, Yet. True Audio ? Woot ? I can listen to music and game at the same time just fine with my GTX 560. lol

Can't wait to upgrade to a whole new rig hopefully I get access to that soon, Will get a 780 and I'll stick with it for years to come.

There are so many great non-proprietary physics engines out there. AMD has just as good 3D performance (within margin of error/variance, like silicon or driver development). AMD's GCN has embraced the open standard of OpenCL. That leaves Shadowplay untouched, but you've already mentioned Mantel and TrueAudio.

[TheBLOODYTOXIC]

In Finland prices are still 399 to 449.

 

haha only greedy americans do the bitcoin mining thing.

[TheMissxu]

btw, there is a reference 290 for $434.99 in stock here (amazon US), and, yesterday, 19 were sold on ebay for $442.99. Edit: As of time of edit, no longer available.

Edited December 24, 2013 by TheMissxu

[Mindtrickz]

My opinion is that 290 might give more perfomanse(icl.Mantle), but 780 will give better effiency, PhysX, better 3D, Cuda, G-sync geforce experience and Shadowplay.

 

 

The thing about Mantle is that, it will be absolutely useless on 90% of the games released. It will only benefit those games that support, and not that many do.

[LAwLz]

I am disappointed with the 290 overclocking. Was expecting it to get at least 15% with water cooling. Guess third party coolers will only fix the eye shattering sound the stock cooler makes.

 

Do we know if they increased the voltage on the cards? The doc doesn't specify it but the text says that if they increased it more it became unstable. Anand's review says their card didn't have any voltage control either. Do PowerTune do that automatically?

[TechFan@ic]

I am disappointed with the 290 overclocking. Was expecting it to get at least 15% with water cooling. Guess third party coolers will only fix the eye shattering sound the stock cooler makes.

 

Do we know if they increased the voltage on the cards? The doc doesn't specify it but the text says that if they increased it more it became unstable. Anand's review says their card didn't have any voltage control either. Do PowerTune do that automatically?

The doc indicates that the card was not over-volted.

The overclocking utility within Catalyst does not include over-volting and according to Linus that's what they used to overclock the 290, so it was definitely not over-volted which is the whole point of putting it under water.

Nvidia does not include support for over-clocking within the Nvidia control panel, so Luke uses EVGA PrecisionX to overclock the Nvidia cards.

A happy middle would be to use MSI Afterburner which supports overclocking & over-volting both AMD & Nvidia GPUs.

In any case Anandtech published their review of the custom R9 290 from sapphire which had a 100% reference PCB and the card was over-volted to 1.23v and overclocked to 1125mhz.

 

[Grave]

He's said two things in previous videos and in this one. Not every card is the exact same, you will have variations when you go to OC. And he said that they will always push their cards to their limit with OC.

 

I still think this was a reasonable outcome, even if his 290 wasn't a decent overclocker/the 780 was a really good one in comparison. However, I have to say that his previous benches for the 290, 780, etc showed that the 780 was better than the 290 on air as well. So, if you really thought that his R9 290 and his GTX 780 would actually perform differently in comparison to one-another, I would say that is very dumb. I think it was a good representation of what a real water block will do to either GPU. It also helps with putting overclocking results into a real scenario. He shows what his cards achieved, and not every card will do that (better or worse), which is what you should be keeping in mind when even considering a custom cooling solution + OC.

 

If the two cards were actually completely different cards from previous benchmarks he's shown on his channel, then you can ignore what I said. And really, this just shows that you don't need massive investments to get decent results out of your cards. I think a lot of people watched this and saw that air isn't total crap on the hottest cards on the planet, at least when a budget is in mind.

[TheMissxu]

Tom's Hardware used an Arctic's Accelero Xtreme III and received a 40MHz higher overclock than your WC 290.

[Mindtrickz]

In any case Anandtech published their review of the custom R9 290 from sapphire and the card out-performed the GTX 780 in all games tested except in BF3 where it was a tie, the card also had unlocked voltage control and was over-volted to 1.23v and overclocked to 1125mhz.

 

You do realize that they tested reference models vs a non-reference model R9 290. Anyway, they didn't overvolt the 290 and that 780 they had was a dud.

[kreotis]

Disappointed in the lack of overvolting. Those cards both can do better on air when overvolted without running into cooling issues.

[TechFan@ic]

You do realize that they tested reference models vs a non-reference model R9 290. Anyway, they didn't overvolt the 290 and that 780 they had was a dud.

It has a 100% reference PCB, the only difference between it and the reference R9 290 is the cooler.

So essentially it's the same as Linus's water-cooled R9 290.

And no, Linus's 780 is not a dud, not at all, it overclocked as well as an over-volted 780 Lightning, which was the best overclocker out of 7 cards in TechPowerUp's testing.

http://www.techpowerup.com/reviews/MSI/GTX_780_Lightning/29.html

[Mindtrickz]

It has a 100% reference PCB, the only difference between it and the reference R9 290 is the cooler.

So essentially it's the same as Linus's water-cooled R9 290.

And no, Linus's 780 is not a dud, not at all, it overclocked as well as an over-volted 780 Lightning, which was the best overclocker out of 7 cards in TechPowerUp's testing.

http://www.techpowerup.com/reviews/MSI/GTX_780_Lightning/29.html

 

The cooler makes a difference considering the GPU Boost. This is why reference model R9 290 doesn't beat the reference model 780. But non-reference model 290 beats reference 780.

 

Hmm.. Maybe I just got a good over-clocker because mine does better than all of those.

[Deeyar]

yo which test bench is that? I want one

[OutlawedConsole]

im supprised about the 780. 

 

i honestly was beliveing the "wait till they watercool the 290" hype

yeah, I was a little bit disssapointed, but then again I'm not because I have a 780

[TheRoss0411]

I just got a 780 and it's perfect

[FellTheSky]

i want this tests but add the 780 ti and r9 290x on the test bench. Also maybe 2 more games?... 

[ukaussi]

At the end of the day all the smart people will fully research both options and make a choice based upon the games they play or will be playing as well as current pricing.

 

I am still sitting on the fence waiing to replace my GTX670 with 2 of either the 290 or 780 with custom coolers and I won't until ASUS etc bring them out so I can get multiple reviews on them and make an INFORMED decision.

[Opsilon]

It has a 100% reference PCB, the only difference between it and the reference R9 290 is the cooler.

So essentially it's the same as Linus's water-cooled R9 290.

And no, Linus's 780 is not a dud, not at all, it overclocked as well as an over-volted 780 Lightning, which was the best overclocker out of 7 cards in TechPowerUp's testing.

http://www.techpowerup.com/reviews/MSI/GTX_780_Lightning/29.html

So what?

The 290 and 290x beat or are on par the 780 at stock speeds, overclocking is not a guarantee anyway, i understand free performance, but a card that is binned at the factory is actually a validated "overclock" over the reference spec.

Linus even stated that the 780 could go higher at 1080p, because it requires less power to drive that resolution, so you can overclock based on the resolution since both cards are power limited over that point, if they had more power they would both go higher with the same voltage increases.

 

Stock is the definitive, all else will be luck tied to the silicon lottery, or binning at the factory, 290 is a monster card, it gives great value, and it can take on a higher priced card any day. Stop crying, as time goes we will see less overlocking potential anyway same as for the CPU-s, it's not just the architecture, its the materials used as well. Overvolting can really shorten your cards lifespan so i would take the 290 stock and oc with what i have over risking a dead card!!!

[TechFan@ic]

So what?

The 290 and 290x beat or are on par the 780 at stock speeds, overclocking is not a guarantee anyway, i understand free performance, but a card that is binned at the factory is actually a validated "overclock" over the reference spec.

Linus even stated that the 780 could go higher at 1080p, because it requires less power to drive that resolution, so you can overclock based on the resolution since both cards are power limited over that point, if they had more power they would both go higher with the same voltage increases.

 

Stock is the definitive, all else will be luck tied to the silicon lottery, or binning at the factory, 290 is a monster card, it gives great value, and it can take on a higher priced card any day. Stop crying, as time goes we will see less overlocking potential anyway same as for the CPU-s, it's not just the architecture, its the materials used as well. Overvolting can really shorten your cards lifespan so i would take the 290 stock and oc with what i have over risking a dead card!!!

That's all well and true, it doesn't change the fact that the testing was unfair because one part (780) was over-volted while the other part (290) was not.

You either overclock both parts without over-volting, overclock both parts with over-volting or don't overclock at all otherwise the results will be unfair.

[TheMissxu]

You realize an aftermarket 290x is being compared with a 780Ti... Not a 780. 
 

Like the ASUS option, the Sapphire card performed better than the GTX 780 Ti in Bioshock Infinite and Crysis 3 while the GTX 780 Ti's only definitive victory came in Battlefield 3.  In the three other games tested, the cards were so close that I'll call it a performance tie.

http://www.pcper.com/reviews/Graphics-Cards/Sapphire-Radeon-R9-290X-Tri-X-4GB-Graphics-Card-Review/Overclocking-and-Concl
 
By that satement, I'd say a stock non-reference 290x beats a stock reference 780Ti (the fact that it's non- vs reference is negligible, because we all know Nvidia's cooler is awesome. No overclocking is more relevant, but PCPer was able to achieve an effective 14% overclock with their 290x).

Edit: Plus I just realized this is pretty irrelevant to the discussion titled, "R9 290 vs 780." Oh well...

Edited January 3, 2014 by TheMissxu

[Opsilon]

That's all well and true, it doesn't change the fact that the testing was unfair because one part (780) was over-volted while the other part (290) was not.

You either overclock both parts without over-volting, overclock both parts with over-volting or don't overclock at all otherwise the results will be unfair.

LTT OC-es all the card and only now it's unfair, as Linus said we have to take everything with a grain of salt, not everyones gpu will oc to the same lvl, so giving OC benchmarks only is unfair to begin with, i know that overvolting should be done for both cards, and that it is unfair, but it does not matter at all, we know how powerful and "cheap" the 290 is, and as a value option and stock performance is on par with the card for less money.

 

I am not only speaking to you, i am speaking to everyone in this thread, chill out, we don't need to be keyboard warriors all the time, when we already know the truth about this entire situation!

[zip2k4]

LTT OC-es all the card and only now it's unfair, as Linus said we have to take everything with a grain of salt, not everyones gpu will oc to the same lvl, so giving OC benchmarks only is unfair to begin with, i know that overvolting should be done for both cards, and that it is unfair, but it does not matter at all, we know how powerful and "cheap" the 290 is, and as a value option and stock performance is on par with the card for less money.

 

I am not only speaking to you, i am speaking to everyone in this thread, chill out, we don't need to be keyboard warriors all the time, when we already know the truth about this entire situation!

 

his point is the test is unfair and it really isn't fair, that's it, no keyboard warrior, no mouse paladins, no monitor overseer.