290X Vs 780 Ti Under Water Performance?

[Z15CAM]

In Canadian dollars, the Ref R9 290X costs approx $600 with Games Vs the Ref GTX 780 Ti for approx $750 with games. Difference is $150 for cards with approx the same performance favoring the 780 Ti in reference dress.

The 290X is 512Bit/4GB vRam Vs the 780 Ti's 384Bit/3GB Vram.

I'm presently building a Water Block 290X costing me approx an additional $750 for a 2 loop system so add that to the price of the 780 Ti to WB it. Still a $150 price difference.

I know a Reference R9 290X can run at a 1195/1500 Mhz clock with a +156mv off-set without throttling with the Fan screaming. I have no clue how fast a Reference GTX 780 Ti will clock without throttling but I imagine it's higher.

Hopefully the 290X will clock somewhere around 1300/1600 to say 1325/1625 Mhz with whatever mv off-set is required under water.

I believe the R9 290X will benefit substantially more then the GTX 780 Ti under water and take the performance lead by a fair amount.

I know we have not much data to compare the performance of these 2 cards with under water as they are just too new but there should be very interesting competition here.

What are your projections?

 

[BallaTheFeared]

Hopefully the 290X will clock somewhere around 1300/1600 to say 1325/1625 Mhz with whatever mv off-set is required under water.

Maybe in a benchmark with 1.4v on a top shelf chip.

 

[Z15CAM]

You just reminded me that Ref 290X 1195/1500 Mhz clocking was done with a +156 mv offset. It will do a 1170/1500 with a +100 mv off set. So I corrected it. Anything higher the Heat builds, card throttles and kills the OC'g.

 

[blackened23]

Personally the 290X has not demonstrated any appreciable REAL WORLD overclocking in anything other than 3 minute suicide valley runs. Until a website can overclock a card for 24/7 stability in actual games, I'd say the nod goes to the 780ti, and is heavily skewed towards the 780ti.

The only overclocks on review websites show the 290X with an average overclock of 1100 for 24/7 stability in actual real world games, with the card gaining around 8% at that. The 780ti is gaining 20-22% over the stock 780ti per reviews with aftermarket cards. Keep in mind the 780ti is doing this on reference PCB cards, on air, with no overvoltage... now with lightning 780ti cards with over-voltage, those cards will gain even more than the 20% than reference PCB cards are getting. The OC scaling with the 780ti is insanely good.

Hopefully the aftermarket 290X cards will show what they're all about but you asked - and I don't think the 290X will match the 780ti under water. I mean, this is the 780ti overclocked on air:

Now with a classified 780ti under water? It will be a good margin faster than even this.

That said, we don't have 290X aftermarket data and those cards could be significantly better in terms of overclocking - the non reference PCB cards should help quite a bit, even with water. It seems that current 290X cards under water are being limited by the reference PCB. So, we'll see.

 

[blackened23]

Despite all that, an aftermarket/water cooled 290X card will clearly be more bang for the buck. While I don't think a 290X card under water will match a 780ti under water (with the above scaling in mind), that's perfectly fine because the card is significantly cheaper.

Heck, if an watercooled non ref 290X costs 560-570$ (excluding the loop) and can meet 90% of the 780ti potential and without the variance/noise of the reference cooler? The card would be great based on that alone. Because a 780ti costs way more money. The 780ti will be faster, but it wont' be 150$ faster.

 

[Z15CAM]

Keep in mind that graph only shows the 290X at 1000/1250 in Umber mode. It will do an easy 1130/1500 with 0 mv off-set without heat issues or throttling running one bench continuously one after the other delivering consistent scores.

I've no clue why they included the Silent Mode as anyone buying these cards never use it.

Since they are not Voltage Locked it really doesn't matter which Mode the Fan Profile is in when under water and MSI Afterburner will over ride it and run the users Fan Profile.

I've found that cooling the GPU is a none issue with the 290X but the VRM's is another matter if you don't have a solid contact copper Heat Sink with adequate cooling such as an Active VRM Water Cooling. That's why the Reference Cooler does a superior job cooling the 290X then the Accelero Xtreme III for OC'g.

 

[ICDP]

Personally the 290X has not demonstrated any appreciable REAL WORLD overclocking in anything other than 3 minute suicide valley runs. Until a website can overclock a card for 24/7 stability in actual games, I'd say the nod goes to the 780ti, and is heavily skewed towards the 780ti.

The only overclocks on review websites show the 290X with an average overclock of 1100 for 24/7 stability in actual real world games, with the card gaining around 8% at that. The 780ti is gaining 35-38% over the stock 780 per reviews with aftermarket cards. Valley is a cool game and all, but isn't representative of a real world overclock - hopefully the aftermarket 290X cards will show what they're all about but you asked - and I don't think the 290X will match the 780ti under water. I mean, this is the 780ti overclocked on air:

Now with a classified 780ti under water? It will be a good margin faster than even this.

That said, we don't have 290X aftermarket data and those cards could be significantly better in terms of overclocking - the non reference PCB cards should help quite a bit, even with water. It seems that current 290X cards under water are being limited by the reference PCB. So, we'll see.

I agree a GTX780 Ti will have superior OC headroom but that chart is making the R9 290/X look worse than it is. Strange that on HardOCP the R9 290X is actually faster at higher res and settings than the results from that HW Canucks chart.

http://www.hardocp.com/article/2013/11/07/nvidia_geforce_gtx_780_ti_video_card_review/3#.UqXFKXlFA-U

I have tested a friends R9 290X CF rig and each card would do 1150-1180 max OC with 100MV. RAM would not clock more than 150MHz and IMHO even though the overall 13%-17% real world performance increase from the OC was nothing spectacular it was certainly decent.

Of course running these OCs was not an option as the cards are far too loud and hot with that absolutely atrocious cooler. During testing his wife came into the room to find out why we were using her hair dryer

Thankfully for the sanity of his family, he is planning to watercool them. I am slightly jealous to be honest, his CF R9 290s unlocked to R9 290X for £600. If it wasn't for that cooler

 

[blackened23]

Yeah, that's true, there just isn't much data right now in terms of maximum 290X overclocks. When the aftermarket 290X cards hit that will change. So I believe that overclocks are really being limited by the reference PCB right now, and not a lot of websites are doing water cooled OC reviews.

So I expect that we'll have way more data in the coming weeks. I could be entirely wrong and the 290X could scale insanely well with non reference PCBs under water. Things should get interesting in the coming weeks, and hopefully we'll get some max OC vs max OC reviews at a few more websites...

 

[Silverforce11]

If you want water results you have to dig it out of OC forums where enthusiasts do those setups. From my own reading, R290/X seem to hit around 1.25ghz only even on water. The reference VRM setup just doesn't liked being pushed further.

 

[ICDP]

If you want water results you have to dig it out of OC forums where enthusiasts do those setups. From my own reading, R290/X seem to hit around 1.25ghz only even on water. The reference VRM setup just doesn't liked being pushed further.

While it is OT (slightly) when I work out the performance increase these unlocked R9 290s got from stock it is very impressive.

R9 290 stock 947/1250

R9 290 OC 1170 = 24% core
Unlock to R9 290X = 5%
An extra 12% for VRAM

Overall this unlock/overclock is a giving ~ 25% extra real gaming performance.

 

[Malik2981]

I agree with the other. The GTX 780 Ti would be the much better investment.

 

[parvadomus]

Dont know what happened in that 780TI review, but the difference with 290X is not that far. In fact stock vs stock, the difference is so slim that I think there´s something wrong with it.
Example, one of the latest reviews: http://www.hardwareluxx.de/index.php/artikel/hardware/grafikkarten/28690-evga-geforce-gtx-780-ti-superclocked-mit-acx-kuehler-im-test.html?start=11

 

[Silverforce11]

Dont know what happened in that 780TI review, but the difference with 290X is not that far. In fact stock vs stock, the difference is so slim that I think there´s something wrong with it.
Example, one of the latest reviews: http://www.hardwareluxx.de/index.php/artikel/hardware/grafikkarten/28690-evga-geforce-gtx-780-ti-superclocked-mit-acx-kuehler-im-test.html?start=11

It really comes down the game selection with these reviews, a few games that prefer AMD or NV and the result is completely skewed.

 

[Z15CAM]

The only AIB's I know of that modify the Reference AMD PCB's are MSI Lightning, Sapphire Toxic and the Asus Matrix versions.

Relocating and adding more VRM's would improve the heat characteristics of the Reference PCB design and allow greater Voltage off-sets.

I've run Valley with the Reference Cooler Screaming at 95% with a +156mv off-set at 1300/1625 Mhz but the GPU heat pegged at 95C and the card throttled killing the OC - Oddly enough the VRM temp didn't go above 76C - That may have been due to throttling though. I did however pull off consistent consecutive 2976 Valley scores, with the CPU idling at 1600 Mhz, with the Ref card clocked at 1195/1500 and a +156mv off-set.

Sure hope water cooling will give me a consistent 1250 to 1300+ / 1550 to 1625 Mhz clock say with somewhere around a 1.44mv off-set.

That should give the 780 Ti a run for it.

It really comes down the game selection with these reviews, a few games that prefer AMD or NV and the result is completely skewed.

Very true. I think members here in Anand Tech Forums would present a better comparison between the two once more of us are setup with these cards.

 

[Will Robinson]

With a GB more VRAM,True Audio and with Mantle still to come the 290X is the better choice.

 

[BallaTheFeared]

With a GB more VRAM,True Audio and with Mantle still to come the 290X is the better choice.

Based on 0 games supporting it currently, makes sense.

 

[Z15CAM]

Geez Balla - You certainly throw in a chuckles with your entrances - LOL

 

[Gloomy]

Based on 0 games supporting it currently, makes sense.

Didn't EA announce 15 games with Mantle support?

 

[raghu78]

Didn't EA announce 15 games with Mantle support?

even if EA announced it won't matter to few who will keep arguing that the games have not yet launched. when the games are launched there will be other arguments like Unreal Engine 4 is more popular than Frostbite 3 and since Unreal Engine 4 doesn't yet support Mantle it does not matter. :biggrin:

OP if you want best bang for buck get a R9 290. watercool it and push it to the max. 1250 -1300 mhz should be possible with watercooling.

 

[blackened23]

Didn't EA announce 15 games with Mantle support?

No. AMD claimed that frostbite 3 would support all games using that engine, and those supposed 15 games are running through 2016 at EA. EA is no longer using UE, and they will be frostbite 3 exclusive....so....the claim was that all of these frostbite 3 games would have Mantle. Apparently, BF4 is getting mantle in December (hopefully DICE can make the game stop crashing as well) and the new NFS game is already using frostbite 3, yet has a 30 fps limit hard coded in and doesn't have mantle. AMD claimed that this title would have Mantle. It will not. So AMD's claims don't necessarily meet reality here.

As far as i'm aware, of these 15 titles 2 of them are being released in 2014. There's the new Dragon Age game and the Star wars game.

And true audio is the most cringe worthy eye rolling feature i've ever seen added to a GPU. Mantle is at least potentially interesting. True audio? Who cares...

 

[BallaTheFeared]

None of which do, and no known gameplay, performance, or any tangible; measurable info has been provided yet.

 

[Z15CAM]

Here are the 2 Official Cards in question Owners Forums that we can reference:

The Official R9 290X Owners Forum: http://www.overclock.net/t/1436497/official-amd-r9-290x-290-owners-club

The Official GTX 780 Ti Owners Forum: http://www.overclock.net/t/1438886/official-nvidia-gtx-780-ti-owners-club

 

[blackened23]

2 minute suicide runs that crash as soon as 3dmark is finished tabulating a score are worthless in determining real world overclocks...

Isn't that all they do at OCN? That's my impression....where are the real world 24/7 stable gaming benchmarks...

 

[Z15CAM]

where are the real world 24/7 stable gaming benchmarks

That amounts to honesty - Personally ( for my own stability) I don't Post Benches that Artifact and they must run consecutively for at least an hour.

On the other hand if I see something that appears unrealistic I may attempt to match them and say it was full of artifacts and was lucky to get through it before it blacked screened.

I'm itching waiting for my Water Solution to arrive in the mail and once set up - Pin my i7 2700K at 4.7 Ghz (May be at 5 GHz's) then play with this 290X under water.

All I can say is 780 Ti's owners beware - After all the RED Team has been the Underdog for a long time - LOL

But I also want 780 Ti owners to respond so that we can make a just comparison.

 

[raghu78]

No. AMD claimed that frostbite 3 would support all games using that engine, and those supposed 15 games are running through 2016 at EA. EA is no longer using UE, and they will be frostbite 3 exclusive....so....the claim was that all of these frostbite 3 games would have Mantle. Apparently, BF4 is getting mantle in December (hopefully DICE can make the game stop crashing as well) and the new NFS game is already using frostbite 3, yet has a 30 fps limit hard coded in and doesn't have mantle. AMD claimed that this title would have Mantle. It will not. So AMD's claims don't necessarily meet reality here.

As far as i'm aware, of these 15 titles 2 of them are being released in 2014. There's the new Dragon Age game and the Star wars game.

And true audio is the most cringe worthy eye rolling feature i've ever seen added to a GPU. Mantle is at least potentially interesting. True audio? Who cares...

Johan Andersson (Technical Director and Frostbite engine lead)of DICE stated in his keynote at APU Summit 2013 that all of EA games will use Frostbite 3. Mantle support will be done once the Frostbite 3 engine gets it for BF4.

http://www.youtube.com/watch?v=N_6CAneoW-0

future games like plants and zombies will support Mantle out of the box (see video at 32:00). all EA games going forward support Mantle because once Frostbite 3 integrates Mantle support its a done deal for all of EA games.