[PcLab] Variety of 28nm GPU tests 2012-2015

[showb1z]

In terms of Features means more Vram, better overclocking, better performance per watt, Gameworks, physX, Day 1 game driver support, shadowplay which far better then any AMD solution.

Haha, that's a good one. What an awesome feature, neutering your performance for disproportional IQ improvements. That's not exclusive to Nvidia hardware though.

If the point of this thread was to claim that a 780(ti) was a better buy than a 290(x) that's ridiculous. The gap will only continue to grow.
They say hindsight is always 20/20, but you're proving that wrong.

 

[ultron]

At the end of 2013, most of the experts(!) suggested GTX 780 Ti over R9 290X. They claimed 780 Ti is stronger GPU. But they didn't think that NVIDIA's driver performance boost support wouldn't last long. Now we see the results:

 

[Creig]

Far better then claims like overclockers dream, world fastest card and 40% faster then Sandy bridge. Both Intel and Nvidia never made any similar claims like these.

*cough* 64 ROPs/2MB L2 cache GTX 970 *cough*

 

[xthetenth]

Hey Desprado, you really want to know why the professional overclockers recommend things? The first word in that should be a big old hint.

But yeah, the in house employee overclocker of a company that only sells NV cards recommends NV cards. That sure is a statement on the quality of NV cards rather than what he's hired to put his name on.

*cough* 64 ROPs/2MB L2 cache GTX 970 *cough*

I'm holding a Pascal in my hands.

Maxwell supports Asynchronous Compute.

(I'm still livid about the 970 one)

 

[Azix]

I think there's a serious misconception regarding who buys expensive GPUs. Just seems so wrong. Just because its a PC component does not mean most of the people interested are knowledgeable and overclocking. A lot will buy it because its what they are told is the best and they have the money. Some will buy it so they do not have to think about overclocking. etc etc etc etc. You are probably more likely to overclock a lower end card to get more performance than a top end one.

 

[xthetenth]

High end GPUs are for price inelastic consumers who want more performance and aren't necessarily any more tech savvy. I've seen a whole bunch of 980 Ti buyers asking for some hand holding while overclocking.

 

[Goatsecks]

At the end of 2013, most of the experts(!) suggested GTX 780 Ti over R9 290X. They claimed 780 Ti is stronger GPU. But they didn't think that NVIDIA's driver performance boost support wouldn't last long. Now we see the results:

I do not know why the 780ti gets so much grief. From the same review you linked and at 1080p, the more appropriate resolution for this gpu, it actually beats the 290X on average by 2% whilst using 48% less power (taken from typical gaming consumption).

Retrospectively we know that the performance tables turn. But this does not make benchmarks from older games any less relevant.

 

[IllogicalGlory]

I do not know why the 780ti gets so much grief. From the same review you linked and at 1080p, the more appropriate resolution for this gpu, it actually beats the 290X on average by 2% whilst using 48% less power (taken from typical gaming consumption).

Retrospectively we know that the performance tables turn. But this does not make benchmarks from older games any less relevant.

That power consumption number is obviously erroneous. They accidentally listed 295X2 numbers for the 290X.

 

[Azix]

I do not know why the 780ti gets so much grief. From the same review you linked and at 1080p, the more appropriate resolution for this gpu, it actually beats the 290X on average by 2% whilst using 48% less power (taken from typical gaming consumption).

Retrospectively we know that the performance tables turn. But this does not make benchmarks from older games any less relevant.

actually, despite what nvidia did to get kepler using less power, it was bigger than the 290x and similar to it in power consumption. its a loss all round. AMD should have taken that round easy but nope.

http://www.tomshardware.com/reviews/geforce-gtx-780-ti-review-benchmarks,3663-17.html

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_780_Ti/25.html

 

[crisium]

780 Ti just milked it's price a bit too high relative to the 290X, but with early reference 290X being a disaster and mining inflation later, I too think it gets too bad of a rep since there were periods where buying a 290X wasn't really tenable. I think the lower Kepler cards are where the performance is truly disappointing, although you have to isolate games from November 2014 to present only to really make it show. Trying to ride out 770 and below to Polaris or Pascal is generally harder than on the 280X and lower.

 

[zlejedi]

At the end of 2013, most of the experts(!) suggested GTX 780 Ti over R9 290X. They claimed 780 Ti is stronger GPU. But they didn't think that NVIDIA's driver performance boost support wouldn't last long. Now we see the results:

LoL at comparing average results taken from two different tests with different games included.

2013 - all games designed for 2006 gpus in PS3/X360
2015 - titles designed for GCN based PS4

Yep it must be drivers that are responsible that GCN gpu has relatively better results in 2015.

 

[littleg]

I think there's a serious misconception regarding who buys expensive GPUs. Just seems so wrong. Just because its a PC component does not mean most of the people interested are knowledgeable and overclocking. A lot will buy it because its what they are told is the best and they have the money. Some will buy it so they do not have to think about overclocking. etc etc etc etc. You are probably more likely to overclock a lower end card to get more performance than a top end one.

It's not just the end user you need to consider with overclocking headroom though, there's the OEMs too. Look at the 980ti and all the factory OC models that are out and absolutely kicking ass.

It makes the product much more desirable for a number of reasons to both OEMs and consumers.

 

[Goatsecks]

actually, despite what nvidia did to get kepler using less power, it was bigger than the 290x and similar to it in power consumption. its a loss all round. AMD should have taken that round easy but nope.

http://www.tomshardware.com/reviews/geforce-gtx-780-ti-review-benchmarks,3663-17.html

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_780_Ti/25.html

I didn't realise it was a dodgy result. Sorry. However, looking at the performance at 1080p, and revised efficiency results, the cards look pretty equal to me. Aside from price. Which may be due to differences in percieved quality issues due to the 290x launch. I wonder how this battle would have fared if custom coolers for the 290x were available at launch?

 

[Silverforce11]

I wonder how this battle would have fared if custom coolers for the 290x were available at launch?

That was a fatal mistake, launching that card with the reference cooler killed an engineering victory.

Hawaii is smaller than GK110 by far and in it's non-throttling custom format, it was actually on par with the bigger 780Ti & Titan. Custom R290X maintain 1ghz and ran ~10% faster than the reference cards.

That and it could actually OC decently with vcore support early on.

We were seeing CF R290X destroy SLI 780Ti at 1440p and 4K back then. But it didn't matter, the hot, loud, power hungry and throttling stigma stuck.

I mean look at the 390/X when it was release. They were identical to custom 290/X.

290/X were a tough sell at $200/$250.. and now people see how good a deal is for the 380/X390/X vs 960/970/980. That's with old GCN tech against new Maxwell... all it needed was a decent cooler.

The good news is I guess they learnt their punishing lesson. The new Wraith cooler is a great move, cool & quiet stock CPU cooler, these are very important metrics. For Polaris, the reference design needs to be excellent. Launch reviews stick, first impressions lasts, all that jazz.

 

[AtenRa]

That was a fatal mistake, launching that card with the reference cooler killed an engineering victory.

Hawaii is smaller than GK110 by far and in it's non-throttling custom format, it was actually on par with the bigger 780Ti & Titan. Custom R290X maintain 1ghz and ran ~10% faster than the reference cards.

That and it could actually OC decently with vcore support early on.

We were seeing CF R290X destroy SLI 780Ti at 1440p and 4K back then. But it didn't matter, the hot, loud, power hungry and throttling stigma stuck.

I mean look at the 390/X when it was release. They were identical to custom 290/X.

290/X were a tough sell at $200/$250.. and now people see how good a deal is for the 380/X390/X vs 960/970/980. That's with old GCN tech against new Maxwell... all it needed was a decent cooler.

The good news is I guess they learnt their punishing lesson. The new Wraith cooler is a great move, cool & quiet stock CPU cooler, these are very important metrics. For Polaris, the reference design needs to be excellent. Launch reviews stick, first impressions lasts, all that jazz.

Agreed, Hawaii still is one of the best GPU designs of the last 3-4 years. It will become even faster than Maxwell (GTX980) in DX-12 games. It was a huge mistake to sent those default cooling cards to reviewers.
You reap what you sow, I believe they learned their lesson if we look at Fiji and wraith.

 

[Innokentij]

Comparing a stock 290X that would throttle in original design vs a much better stock design on 780TI it would get stomped. There has been no driver downgrade of performance and this has been tested online just click the link under and look at the 3dmark results. 780TI was build for it's time and delivered everything u wanted for the price, it gave u more then titan performance for 300 dollars less and was the true champion. 3 year later some noobs still fighthing for 2013 performance crown with a 780TI none OC stock cooler vs a OC'ed 3rd party 290X with benchmarks they dont even understand and dont understand what maxwell architecture brought to the table. No words.

Read and educate urself:

http://www.bytemedev.com/the-gtx-780-ti-sli-end-of-life-driver-performance-analysis/

Here is some best case for 290X, imagine how much worser it did with the stock cooler.

Source : http://www.hardocp.com/article/2014/04/14/gigabyte_r9_290x_oc_asus_gtx_780_ti_dcuii_review

 

[Azix]

Comparing a stock 290X that would throttle in original design vs a much better stock design on 780TI it would get stomped. There has been no driver downgrade of performance and this has been tested online just click the link under and look at the 3dmark results. 780TI was build for it's time and delivered everything u wanted for the price, it gave u more then titan performance for 300 dollars less and was the true champion. 3 year later some noobs still fighthing for 2013 performance crown with a 780TI none OC stock cooler vs a OC'ed 3rd party 290X with benchmarks they dont even understand and dont understand what maxwell architecture brought to the table. No words.

Read and educate urself:

http://www.bytemedev.com/the-gtx-780-ti-sli-end-of-life-driver-performance-analysis/

Here is some best case for 290X, imagine how much worser it did with the stock cooler.

http://www.hardocp.com/images/articles/1397411267OPl8cM2MpM_5_3_l.jpg
[IMG]http://www.hardocp.com/images/articles/1397411267OPl8cM2MpM_6_3_l.jpg[IMG]http://www.hardocp.com/images/articles/1397411267OPl8cM2MpM_7_3_l.jpg
[IMG]http://www.hardocp.com/images/articles/1397411267OPl8cM2MpM_8_3_l.jpg

Source : [url]http://www.hardocp.com/article/2014/04/14/gigabyte_r9_290x_oc_asus_gtx_780_ti_dcuii_review[/url][/QUOTE]

That all seems kind of pointless. The issue is not with firestrike or older games. Not sure if the people saying downgrade really do mean they are sabotaging with driver updates, but my thinking was always that its getting left behind. What ultimately matters is performance in newer games and that has been sucking for kepler since last year at least.

I doubt they are going to go back in drivers to change performance in firestrike and oldler games. For games they COULD but I doubt they would go that far. Just let kepler languish a bit. They can optimize for kepler. Its not a worlds difference

 

[AtenRa]

That all seems kind of pointless. The issue is not with firestrike or older games. Not sure if the people saying downgrade really do mean they are sabotaging with driver updates, but my thinking was always that its getting left behind. What ultimately matters is performance in newer games and that has been sucking for kepler since last year at least.

Yea that is exactly the problem with Kepler, it run out of steam real quick. And im sure Maxwell will start to run out of steam the moment DX-12 games will start to popup, especially when Pascal will be released.

 

[Innokentij]

Yea that is exactly the problem with Kepler, it run out of steam real quick. And im sure Maxwell will start to run out of steam the moment DX-12 games will start to popup, especially when Pascal will be released.

But it havent that is the point, nobody is using a stock thermal throttling 290X in the overview of the performance anymore. And all 3rd party 290X is OC'ed. It just looks like it has fallen behind when it really hasent it's just not competing on the same terms and this is why i never use this graphs or take em seriouse, want to show difference in performance show us what card u use, show us graph of time with min average max fps, show us FCAT, show us mhz stats. Im really feed up with below my standards review sites that pumps out mass cards but dont go into detail about each card used. Dont have to have it in the review of the game like say on game gpu but atleast have a click here to check what cards was tested and what stats was set on em.

 

[AtenRa]

But it havent that is the point, nobody is using a stock thermal throttling 290X in the overview of the performance anymore. And all 3rd party 290X is OC'ed. It just looks like it has fallen behind when it really hasent it's just not competing on the same terms and this is why i never use this graphs or take em seriouse, want to show difference in performance show us what card u use, show us graph of time with min average max fps, show us FCAT, show us mhz stats. Im really feed up with below my standards review sites that pumps out mass cards but dont go into detail about each card used. Dont have to have it in the review of the game like say on game gpu but atleast have a click here to check what cards was tested and what stats was set on em.

Im not only talking about GTX780Ti, the entire Kepler line has run out of steam since 2014-15. Take the GTX680 or even GTX770 and compare it to HD7970 or R9 280X.
Now days the GTX770 is slower even than the cut off Tonga R9 285 witch is almost the same as the HD7950 GHz.