[H] GTX 780 Ti vs. R9 290X 4K Gaming

[Ferzerp]

Amusingly, because I've come to use a shield in console mode so much, I have exactly 0 incentive to upgrade to either of these solutions.

 

[bystander36]

What a joke of a cherrypicked post, good work. The green team here praised H when Nvidia "felt smoother". Hypocrites.


Enough of the attacks, please.

-Rvenger

It is quite clear by all the reviews at 4k resolutions (perhaps other resolutions too), that AMD handles higher levels of AA without as much loss of performance. However, those settings are also unplayable at 4k.

Hardocp's results, show the 780ti as outperforming the 290x, until they do the apples to apples test with AA cranked up and at unplayable settings.

 

[blackened23]

It is quite clear by all the reviews at 4k resolutions (perhaps other resolutions too), that AMD handles higher levels of AA without as much loss of performance. However, those settings are also unplayable at 4k.

Hardocp's results, show the 780ti as outperforming the 290x, until they do the apples to apples test with AA cranked up and at unplayable settings.

I don't think it's even that. AMD cards have been able to do 2-screen surround gaming for some time and that is precisely what 4k MST is designed as: 2 screen surround. Nvidia cards have never been able to do 2 screen surround gaming - NV has only been able to span 3d apps across 3 screens. So I suppose NV is playing catch up in terms of driver compatibility and what not with 4K resolution / MST being essentially 2 screen surround. That is new territory for NV cards, which have only been able to do 1+1 (no surround) or 3 (surround) with fermi and 1+1 (no surround) or 3 (surround) +1 with Kepler. Conversely, AMD has been able to do 2 (surround) or 1+1 (no surround) or 3 (surround) or 5(surround)+1 for a long period of time.

I'm sure NV will increase performance relatively soon, as NV has always without exception excelled at software. If they can find a way to increase MST surround performance i'm sure they will do precisely that.

 

[Silverforce11]

It is quite clear by all the reviews at 4k resolutions (perhaps other resolutions too), that AMD handles higher levels of AA without as much loss of performance. However, those settings are also unplayable at 4k.

Hardocp's results, show the 780ti as outperforming the 290x, until they do the apples to apples test with AA cranked up and at unplayable settings.

Playable at 4K is still beyond single GPUs, turning down too many settings sacrifices visual, you may as well stick with 1600p and maxing out games.

 

[toyota]

So you are saying the $700 card overclocks to better performance than the $400 card???

of course I meant the 290x if you look at what I am replying to

 

[bystander36]

Playable at 4K is still beyond single GPUs, turning down too many settings sacrifices visual, you may as well stick with 1600p and maxing out games.

The only setting that seemed to have a big effect was AA. At that resolution, the 290X handled it much better. Without AA or lower forms of AA in the non apples to apples test, where they focus on playable settings, they do well.

I'm not really sure you can consider no AA as a reason not to use a 4k screen, as at that resolution, AA is far less needed.

 

[bystander36]

I don't think it's even that. AMD cards have been able to do 2-screen surround gaming for some time and that is precisely what 4k MST is designed as: 2 screen surround. Nvidia cards have never been able to do 2 screen surround gaming - NV has only been able to span 3d apps across 3 screens. So I suppose NV is playing catch up in terms of driver compatibility and what not with 4K resolution / MST being essentially 2 screen surround. That is new territory for NV cards, which have only been able to do 1+1 (no surround) or 3 (surround) with fermi and 1+1 (no surround) or 3 (surround) +1 with Kepler. Conversely, AMD has been able to do 2 (surround) or 1+1 (no surround) or 3 (surround) or 5(surround)+1 for a long period of time.

I'm sure NV will increase performance relatively soon, as NV has always without exception excelled at software. If they can find a way to increase MST surround performance i'm sure they will do precisely that.

That may be part of it. I guess if we look at the 1600p results with AA cranked up and not, we'd see if AA hurts Nvidia cards as much at more mainstream resolutions, where AA will definitely be used.

 

[Pandora's Box]

The only setting that seemed to have a big effect was AA. At that resolution, the 290X handled it much better. Without AA or lower forms of AA in the non apples to apples test, where they focus on playable settings, they do well.

I'm not really sure you can consider no AA as a reason not to use a 4k screen, as at that resolution, AA is far less needed.

People said the same thing about AA and 1440P/1600

 

[OCGuy]

Cool.....although the vast majority of owners of those cards are not playing on 4k....

I just don't see 4k as a viable option until the next node or beyond.

 

[Silverforce11]

Cool.....although the vast majority of owners of those cards are not playing on 4k....

I just don't see 4k as a viable option until the next node or beyond.

CF R290/X run 4K easily even with AA, its certainly viable IF you have the $$ for those monitors.

As soon as a good 4K monitor drops to around $1000, then we'll see much higher uptake.

 

[bystander36]

People said the same thing about AA and 1440P/1600

Short of the unplayable 4k resolution benches, or similar Eyefinity benchmarks, I see more 780ti wins. Of course many people are very focused on the 4k benches. I prefer seeing benchmarks at resolutions and FPS people play at.

 

[OCGuy]

Short of the unplayable 4k resolution benches, or similar Eyefinity benchmarks, I see more 780ti wins. Of course many people are very focused on the 4k benches. I prefer seeing benchmarks at resolutions and FPS people play at.

It is common for a certain metric to be picked out of the bunch and highlighted. A couple years ago it was heat/noise/power. Then when NV put throttling on the 590 because reviewers were intentionally trying to over-volt them until they "blew-up", that was suddenly a big deal. For some reason those are ignored now by the 4k pushers.

780ti is the fastest single-gpu card available until sometime next year, and people will pay a premium for that.

By the way, I didn't see any great detail in the OP about power/heat/noise...?

 

[bystander36]

It is common for a certain metric to be picked out of the bunch and highlighted. A couple years ago it was heat/noise/power. Then when NV put throttling on the 590 because reviewers were intentionally trying to over-volt them until they "blew-up", that was suddenly a big deal. For some reason those are ignored now by the 4k pushers.

780ti is the fastest single-gpu card available until sometime next year, and people will pay a premium for that.

By the way, I didn't see any great detail in the OP about power/heat/noise...?

I just assumed the reason they always try to pick unplayable settings, is to either ensure the CPU won't bottleneck anything, or they just enjoy testing raw power at the highest settings possible.

I much prefer when they choose settings at playable FPS. Those are the ones I focus on.

 

[boxleitnerb]

So basically the 290X and 780 Ti are equal at 4K (since the 290X was again tested in uber mode, the 780 Ti was not changed...). Good to know.

 

[3DVagabond]

Motherboards for the most part do not let you pick and choose which slots to use for SLI or CFX. With asus and MSI motherboards there are only 2 slots to use for maximum speed CFX - SLI, despite having way more PCIE slots. This will be specified in the manual. If you deviate from the recommended 2 slots, you will run at x4/x8.

Well, technically you COULD use different slots but one of the cards would then be running at x4 speed. Which is not ideal for obvious reasons. Generally speaking, most motherboards require 2 specific slots (asus does anyway) for full speed (eg x8/x8 or x16/x16) crossfire or SLI. I don't know if GBT is different in this respect, but most motherboard manuals specify which 2 slots to use for SLI and they're generally too close for comfort in using open air cards. I'm not sure if your gigabyte motherboard is different in this respect.

Like anything else when you aren't just running the basics, you need to make sure your components are compatible for the situation. There are plenty of mobo's that have triple spacing and cases that move plenty of air past the cards. You have to understand when your setting something like this up that you are going to have to dissipate ~500w of heat for the 2 cards and have a setup that's capable of doing that.

Reminds me of all the threads where people had 3x DVI monitors and complained when they found out they needed 1 dp monitor or an active dp>dvi adapter to do it. You need to check those things before you buy the components.

 

[3DVagabond]

GURU3D R9-290X review (where the numbers come from for the 780 Ti 4K comparison)

"For the remainder of the tests we use the card's default settings, BIOS 1 quiet mode."

Pretty shocking that sites would use this to advantage the 780ti in comparisons.

 

[blackened23]

Pretty shocking that sites would use this to advantage the 780ti in comparisons.

I suspect more sites will do this. What's shocking is the fact that AMD asks the user to choose between silence or performance, when the 780ti doesn't. Then again, we could just change everything by tweaking the 780ti's power targets and what not to go against uber mode. Sounds great right? Additionally, you will have users who will just look at pretty FPS graphs, buy a 290x card from a retailer, and then be shocked to realize that their default 290x mode doesn't perform as they had expected. I'm fairly certain that none of the cards actually document the differences between the 2 BIOS', even if AMD does warranty them - i've read of several 290X users being completely oblivious to the "uber mode". Not everyone reads websites like this every day to discern the differences.

To be clear, I don't have an issue with uber mode benchmarks. It's completely up to the reviewer and whether they feel like testing the same card twice and wasting their time, or whether they want to normalize sound levels to reasonable levels between the 290X and 780ti. Regardless, AMD isn't the victim is the website chooses to not test it, fact of the matter is their reference design is terrible (100% AMD's fault) and they shouldn't ask users to compromise like that. IMHO.

What you SHOULD be doing, if you truly care about this issue, is petition AMD to update their low quality reference design. Then you won't have to complain when more websites stop testing uber mode. THAT will fix the problem. Complaining here won't fix it, because like I said. I suspect more websites will follow, why should they waste their time testing a card twice? The default setting is what consumers get and the BIOS differences aren't clearly documented on 290X packaging boxes from what i've heard.

 

[wand3r3r]

By the way, I didn't see any great detail in the OP about power/heat/noise...?

The article didn't highlight the power/heat/noise?

It's designed to run at 95C, no need to "highlight" that. Every thread so far has been full of "noise" complaints (primarily be people who don't own the cards), so I don't see a need to make something up about that. I agree it's probably louder than I'd personally like, however that's subjective, while the review is primarily about objective data such as performance. Interesting as subjective data is, noise is not really the concern and we all know that's about the only advantage the 780 ti has.

Power concerns are also somewhat subjective. I am not concerned, given the alternative is in the same ballpark with both performance and power.

Power = close
Performance = close
Noise = 780 ti wins

The interesting part to me is that the 780 ti doesn't have the performance advantage that NV cards have had such as the 6970-580 gap, yet the price premium is large.

 

[GaiaHunter]

I suspect more sites will do this. What's shocking is the fact that AMD asks the user to choose between silence or performance, when the 780ti doesn't. Then again, we could just change everything by tweaking the 780ti's power targets and what not to go against uber mode. Sounds great right? Additionally, you will have users who will just look at pretty FPS graphs, buy a 290x card from a retailer, and then be shocked to realize that their default 290x mode doesn't perform as they had expected. I'm fairly certain that none of the cards actually document the differences between the 2 BIOS', even if AMD does warranty them - i've read of several 290X users being completely oblivious to the "uber mode". Not everyone reads websites like this every day to discern the differences.

To be clear, I don't have an issue with uber mode benchmarks. It's completely up to the reviewer and whether they feel like testing the same card twice and wasting their time, or whether they want to normalize sound levels to reasonable levels between the 290X and 780ti. Regardless, AMD isn't the victim is the website chooses to not test it, fact of the matter is their reference design is terrible (100% AMD's fault) and they shouldn't ask users to compromise like that. IMHO.

What you SHOULD be doing, if you truly care about this issue, is petition AMD to update their low quality reference design. Then you won't have to complain when more websites stop testing uber mode. THAT will fix the problem. Complaining here won't fix it, because like I said. I suspect more websites will follow, why should they waste their time testing a card twice? The default setting is what consumers get and the BIOS differences aren't clearly documented on 290X packaging boxes from what i've heard.

What I want to know is how this card perform at a given clock rate.
With the uber mode that clock rate seems to be 1 GHz (maybe with a minor variation with a couple of games)
Aftermarket coolers will have no problem sustaining that clock rate.
In silent mode you have a rollercoaster of clock rates.

Basically next time a situation like this happens there will only be uber mode in the card - the 290 is already like that.

 

[blackened23]

What I want to know is how this card perform at a given clock rate.
With the uber mode that clock rate seems to be 1 GHz (maybe with a minor variation with a couple of games)
Aftermarket coolers will have no problem sustaining that clock rate.
In silent mode you have a rollercoaster of clock rates.

Basically next time a situation like this happens there will only be uber mode in the card - the 290 is already like that.

Aftermarket coolers are excessively difficult to use in SLI or xfire, chances are throttling will be a huge issue with any open air cooler in XFIRE. That's essentially why I have a difficult time excusing AMD for re-using their low quality reference design - a lot of people cannot use open air aftermarket cards. It was forgivable on the 5870, 6970 and 7970 because none of those cards lost frequencies due to quiet fan speeds - you could essentially run all of those cards at quiet profiles with no compromises.

 

[GaiaHunter]

Aftermarket coolers are excessively difficult to use in SLI or xfire, chances are throttling will be a huge issue with any open air cooler in XFIRE. That's essentially why I have a difficult time excusing AMD for re-using their low quality reference design - a lot of people cannot use open air aftermarket cards. It was forgivable on the 5870, 6970 and 7970 because none of those cards lost frequencies due to quiet fan speeds - you could essentially run all of those cards at quiet profiles with no compromises.

And what does CF have to do with testing single cards in uber mode?

Also, one can run 290X uber mode in CF.

I'm sorry but 5870, 6970 and 7970 aren't quiet cards, especially in CF - they are less loud.

 

[blackened23]

edit: dp

 

[GaiaHunter]

Are you playing games again? Read the article. SLI AND XFIRE 290X and 780ti 4k PERFORMANCE. What does single card testing in uber mode have to do with the thread?

I think you are in the wrong thread.

The thread title is

[H] GTX 780 Ti vs R9 290X 4K Gaming.

It links to a review with single cards.

 

[blackened23]

Look, playing games over "what does this have to do with" is getting old. Open air coolers aren't for everyone, especially mGPU users. Therefore AMD passing the buck to AIB makers doesn't excuse them, period.

The fact that you think it is okay for AMD's reference design to throttle at quiet fan speeds, I do not get it....why? AMD can and should fix this with a new revision, there is nothing unreasonable about wanting AMD to do just that. You can discuss how you want to see uber mode numbers all day long but that doesn't change the fact that this should be a non issue in the first place - there should not be a "quiet" or "performance" preset. AMD should combine both into one - nothing unreasonable about that.

 

[GaiaHunter]

Look, playing games over "what does this have to do with" is getting old. Open air coolers aren't for everyone. Therefore AMD passing the buck to AIB makers doesn't excuse them. End of story. Period.

The fact that you think it is okay for AMD's reference design to throttle at quiet fan speeds is questionable. AMD can and should fix this with a new revision.

I think that some people care about the noise.
Some people don't care about the noise.
Some people care about saving money.
Other people will prefer to spend more money to have less noise.

Not every consumer out there is blackened23.

In your opinion the R9 290X and the 290 aren't worth it even though they are cheaper because of the noise. We got it the first time.

The reviews include noise charts so people are aware of the noise.