Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
Saladin

How to pick a new GPU; Overview: 780, 780 ti, R9 290x

Recommended Posts

Posted · Original PosterOP

I posted this in another thread (http://linustechtips.com/main/topic/74406-why-is-everyone-hating-on-the-r9-290x/), however it may be useful to people looking to purchase a new GPU so I figured I would re-post it here for the sake of visibility. Below is my methodology for looking at GPUs as well as some interesting data on thermal performance.

 

The two major ways I look at cards are listed below (all numbers are for 1080p as that is the resolution I personally game in).

 

The first is in terms of performance per clock (I'm using OC3D benchmark numbers from Tomb Raider at 1080p though this could be repeated with any benchmarks you want, this was just the first one I clicked and is actually a game I play http://www.overclock...80_ti_review/17) :

  • GTX 780 (clock 863, fps 56.6, fps/clock 0.06559)
  • R9 290x (clock 947, fps 78.5, fps/clock 0.08289)
  • GTX 780 ti (clock 857, fps 87.8, fps/clock 0.10245)
  • GTX 780 ti OC (clock 1101, fps 104.7, fps/clock 0.09509)

As you can see here the 290x sits between the 780 and the 780 ti in terms of performance per clock at 1080p with the 780 ti having a sizable lead. You can also see that there are diminishing returns on clock speed, that is to say as you increase the clock speed, the corresponding fps boost you receive increases at a lower rate.

 

The other way I like to compare cards is in terms of price to performance (same benchmark numbers, cost in USD, ratio shows frames gained per dollar spent) (I also use a formula 100-(1/price to performance) to get a score as this is easier to understand) :

  • GTX 780 (cost $499.99, fps 56.6, performance/price 0.1132) (score 91.116)
  • R9 290x (cost $549.99, fps 78.5, performance/price 0.14273) (score 92.994)
  • GTX 780 ti (cost $699.99, fps 87.8, performance/price 0.1254) (score 92.026)
  • GTX 780 ti OC (cost 699.99, fps 104.7, performance/price 0.1496) (score 93.316)

What these numbers tell you is that the R9 290x has better price to performance than either the stock 780 or the stock 780 ti (higher is better), however the overclocked 780 ti has better price to performance than the 290x. As you can easily see from both metrics it pays to overclock if possible.

 

Here is a comparison of performance per watt, these numbers are not particularly useful for buying a card but do give interesting information about how efficiently the cards use power in producing performance. All numbers are for stock cards not overclocks as the power draw for a given overclock can vary greatly, you will also see variations in power draw between cards even at stock clocks though these should ideally be much smaller.

  • GTX 780 (fps 56.6, TDP 250 watts, performance/watt 0.2264)
  • R9 290x (fps 78.5, TDP 270-300 watts depending on source, performance/watt 0.29074-0.26167)
  • GTX 780 ti (fps 87.8, TDP 250 watts, performance/watt 0.3512)

For a final metric I have included here a performance to temperature ratio, this is a useful way to look at the efficiency of the stock coolers. The data gives you the frames per degree Celsius. Using the same OC3D benchmarks and using their thermal data for these cards; all temps are at load. This could easily be repeated for idle temps, and I will do so if there is a request for it.

  • GTX 780 (fps 56.6, max temp 65, fps/temp 0.8707)
  • R9 290x (fps 78.5, max temp 95C, fps/temp 0.8263)
  • GTX 780 ti (fps 87.8, max temp 71C, fps/temp 1.2366)
  • GTX 780 ti OC (fps 104.7, max temp 79C, fps/temp 1.3253)

As you can see the stock Nvidia cooler is significantly better than the stock AMD cooler; not only is it better out of the box but it scales very well as temperatures increase.

 

I haven't included the overclocked 780 as I could not readily find the OC3D overclock speed for the card they used in their benchmarks, if some one can find them I will update this. I will add the R9 290 (non-x) when/if OC3D reviews it, I do not want to mix benchmarks. I chose to use OC3D because I trust their reviews; the reviews are very in depth with excellent charts. I always recommend doing something similar to this before buying a new GPU, this is not necessarily the best way, but it is how I do it. Hope this is helpful.


Intel 3570k @ 4.4 GHz |Asus Sabertooth Z77 |EVGA GTX 660 Ti FTW |Kingston HyperX Beast 16 Gb DDR3 1866 (2x8Gb)


|Samsung 840 250 GB |Western Digital Green 2TB 2x |Cooler Master 850w 80+ Gold |Custom Water Cooling Loop |Noctua NF-F12 4x
|Noctua NF-A14 3x |Corsair Carbide 500R (White) |Corsair K95 |Razer Mamba |Razer Megalodon |Samsung SyncMaster T220 2x Computer Bucket List   Greatest Thread Ever   WAN Show Drinking Game  GPU Buyers Guide
Link to post
Share on other sites
Posted · Original PosterOP

It is a nice overview but I doubt these calculations will make choosing easier.

Depends on what you look for in a GPU. Personally I like to pick cards with excellent price to performance, using metrics like this quantifies that ratio. This type of approach can be used for any cards, I simply used the 780, 290x and 780 ti as they are the top end single GPU cards on the market. There are of course other things to look for when choosing a card, such as extras (such as games or discounts), warranty, build quality, aesthetics, comparability, and support for specific games. Metrics aren't the only thing to look at, but they can be helpful, especially in cutting through the marketing bs and fan-boying.


Intel 3570k @ 4.4 GHz |Asus Sabertooth Z77 |EVGA GTX 660 Ti FTW |Kingston HyperX Beast 16 Gb DDR3 1866 (2x8Gb)


|Samsung 840 250 GB |Western Digital Green 2TB 2x |Cooler Master 850w 80+ Gold |Custom Water Cooling Loop |Noctua NF-F12 4x
|Noctua NF-A14 3x |Corsair Carbide 500R (White) |Corsair K95 |Razer Mamba |Razer Megalodon |Samsung SyncMaster T220 2x Computer Bucket List   Greatest Thread Ever   WAN Show Drinking Game  GPU Buyers Guide
Link to post
Share on other sites

His charts are pretty crap imo. He needs to add the FPS in with the bar charts, makes reading them difficult with a lower res monitor. Also, how do you benchmark a GTX 780 with a core clock of 863Mhz? Did he disable boost? most if not all GTX 780's will hit 950Mhz are up and stay they for awhile... all of the time with if you change the fan profile.


5820K - ASUS X99-A - 16GB Corsair LPX - HD 7970 GHz - Qnix 1440p @ 96Hz - Waiting for Polaris/Pascal

Link to post
Share on other sites

His charts are pretty crap imo. He needs to add the FPS in with the bar charts, makes reading them difficult with a lower res monitor. Also, how do you benchmark a GTX 780 with a core clock of 863Mhz? Did he disable boost? most if not all GTX 780's will hit 950Mhz are up and stay they for awhile... all of the time with if you change the fan profile.

 

with gpu boost 2.0 alone both my 780s reach 1110 mhz without even oc'ing them, so I feel like different things should be tested :S


Stuff:  i7 7700k @ (dat nibba succ) | ASRock Z170M OC Formula | G.Skill TridentZ 3600 c16 | EKWB 1080 @ 2100 mhz  |  Acer X34 Predator | R4 | EVGA 1000 P2 | 1080mm Radiator Custom Loop | HD800 + Audio-GD NFB-11 | 850 Evo 1TB | 840 Pro 256GB | 3TB WD Blue | 2TB Barracuda

Hwbot: http://hwbot.org/user/lays/ 

FireStrike 980 ti @ 1800 Mhz http://hwbot.org/submission/3183338 http://www.3dmark.com/3dm/11574089

Link to post
Share on other sites
Posted · Original PosterOP

The 780Ti OC doesnt make much sense since you can save $20 and just oc yourself lol.

The metrics are for the same card, one set of data for when it is stock, another for when it is overclocked. As for buying cards that come with an overclock out of the box (ie the 290x) that can be good or bad, sometimes (as with the EVGA 780 SC they are higher binned chips that are still overclock-able). Regardless it pays to overclock your GPU, no reason not to as long as it is stable and the thermals are good.


Intel 3570k @ 4.4 GHz |Asus Sabertooth Z77 |EVGA GTX 660 Ti FTW |Kingston HyperX Beast 16 Gb DDR3 1866 (2x8Gb)


|Samsung 840 250 GB |Western Digital Green 2TB 2x |Cooler Master 850w 80+ Gold |Custom Water Cooling Loop |Noctua NF-F12 4x
|Noctua NF-A14 3x |Corsair Carbide 500R (White) |Corsair K95 |Razer Mamba |Razer Megalodon |Samsung SyncMaster T220 2x Computer Bucket List   Greatest Thread Ever   WAN Show Drinking Game  GPU Buyers Guide
Link to post
Share on other sites
Posted · Original PosterOP

Is the EVGA version of the 780ti SC with ACX cooler good?

I haven't run metrics on that specific card however in general I like cards from EVGA, they have excellent customer service with great warranty support, decent coolers (such as the ACX), they look good, and they offer a step up program (if a better card comes out within 90 days you can exchange yours for the price difference). EVGA SC cards have often been higher binned chips meaning they perform better out of the box and can still overclock. If you have the money to spend, and want to buy a 780ti, EVGA is a good choice. Hope this helps.


Intel 3570k @ 4.4 GHz |Asus Sabertooth Z77 |EVGA GTX 660 Ti FTW |Kingston HyperX Beast 16 Gb DDR3 1866 (2x8Gb)


|Samsung 840 250 GB |Western Digital Green 2TB 2x |Cooler Master 850w 80+ Gold |Custom Water Cooling Loop |Noctua NF-F12 4x
|Noctua NF-A14 3x |Corsair Carbide 500R (White) |Corsair K95 |Razer Mamba |Razer Megalodon |Samsung SyncMaster T220 2x Computer Bucket List   Greatest Thread Ever   WAN Show Drinking Game  GPU Buyers Guide
Link to post
Share on other sites

I haven't run metrics on that specific card however in general I like cards from EVGA, they have excellent customer service with great warranty support, decent coolers (such as the ACX), they look good, and they offer a step up program (if a better card comes out within 90 days you can exchange yours for the price difference). EVGA SC cards have often been higher binned chips meaning they perform better out of the box and can still overclock. If you have the money to spend, and want to buy a 780ti, EVGA is a good choice. Hope this helps.

Thanks so much man :)

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×