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

5700 XT with Image Sharpening & 80% GPU Scaling => 2080 TI performance??

Recommended Posts

Posted · Original PosterOP

5700 XT with Image Sharpening & 80% GPU Scaling => 2080 TI performance!

In my endless wait for 5700XT AIB card, I have weighed a lot of back and forth between 5700 XT and 2070 Super, given that I have a 3440x1440 screen (~ 90hz) and want some future proofing. Something that has not been extensively tested and discussed is Radeon Image Sharpening (RIS) combined with GPU scaling for DX9 / DX12 titles (potentially DX11 if the technology becomes popular). I have tried to gather all the sources I found and try below to try to summarize the results / conclusions of these. TechSpot/Hardware unboxed did the most comprehensive review and Igor have also made a walkthrough of the image sharpening features compared to DLSS (links in the end)

 

Radeon Image Sharpening

  • In comparable tests, RIS wins in almost all games compared to Nvidia's DLSS (possibly the exception of Tomb Raider where DLSS delivers well). Noticeable improvement in sharpness in games without major downsides (some vegetation may be a little too sharp).

  • Nvidia's DLSS only works in combination with Ray Tracing where the GPU is valued, and DLSS steals a lot of performance from the card while this is near intangible for 5700 / 5700XT (max -1.5% performance in FPS drop) A huge advantage for Radeon is that it works for all DX9 and DX12 games without any need for local customization in all games.

GPU Scaling & Radeon Image Sharpening

  • Most sources I linked to below have tested 1440p upscaled to 4k resolution and they think it is a very good tradeoff for close to 50% performance gain, but you can easily see the difference if you zoom in and compare native / scaled resolution. Hardware unboxed tested alot of different gpu scaling options and found the 70-80% scaling sweetspot where its very hard to spot quality differences between scaled and native 4k. 80% scaling provides a 27% FPS boost with hardly no noticable quality loss!

     

5% Overclocking should easily be reasonable for an AIB card

  • Combine this with an AIB card that creates clearly better cooler solutions that should at least create overclocking space of 5% (with only automatic overclocking as I interpret it based on this test: https://www.youtube.com/watch?v=08QgzHFBjq0)

Summary: Overclock + ~ 80% GPU Scaling + RIS gives comparable performance as a 2080 TI card!

  • This gives a base performance of 105% + an additional 27% FPS boost to very small identifiable quality loss! 
    In Sweclockers 5700 XT Test the 5700 XT was tested in 4k with ultra-settings DX12 among a series of games to get a performance average over a multitude of titles. They got a total of 123% performance (compared to GTX 1080). Multiply this by 1.05 for overclocking and you land at 129. If you then use ~ 80% GPU Scaling and RIS giving  you a 27% boost to 164, which is very close to the 2080 TI performance (169) !!

  • 17581?key=0b27fdf7a0b5698f0044213fb69844

Just the fact that this is an on / off button for all DX9 / DX12 games (really hoping for DX11 during the year) is a huge benefit to the NAVI cards (works for other cards too but with a bit more performance loss than <1, 5%). This really makes me await the AIB cards and most likely will run on a 5700 XT. With a little cheating you reach the performance equivalent to a 10,000 SEK card which should give you a fairly strong performance with this card for the foreseable future. 

One trait you lose is of course Ray Tracing, but all cards except possibly 2080TI today find it difficult to deliver a stable 4k experience with ray tracing. For my part, I have no problem waiting until Ray Tracing breaks through before investing in that technology, and then a $ ~450 card is a decent investment until that day.

 

Surprised that no more people have tested this, seeing how striking the results are. I welcome your thought and conclusions

 

I have the following reviews as sources: 
https: //www.youtube.com/watch? V = Yi-_T3vsv-Q & feature = youtu.be & ... 
https://www.youtube.com/watch?time_continue=581&v=7MLr1nijHIo 
https: //www.sweclockers.com/test/27832-amd-radeon-rx-5700-och ... 
https: //www.igorslab.media/ultimativa-visueller-vergleich-zw ...

Link to post
Share on other sites
1 minute ago, Lindstroem said:

given that I have a 3440x1440 screen (~ 90hz) and want some future proofing.

Yeah, I end up going all out on the 2080 Ti at the end of the day, from experience with my 1080 Ti there's the need to fine tune every game to keep on the 100fps margin (also 100hz 1440p ultrawide), considering the 5700XT is ever so slightly slower than the 1080 Ti on average when you OC both I felt that AMD is really out of the picture for me as I wanted to keep as close to ultras as possible.

 

I waited for the 2080 Super but at the end of the day nVidia brought us a card that is 8% faster on average than a 1080 Ti at the same price point, disappointing specially when latest AAA games are even more demanding like Metro Exodus and upcoming RDR2 and Cyberpunk 77... so I just accepted the massive expensive price tag on the 2080 Ti to be safe.

 

Sad part is that I don't give a crap to Ray Tracing, it doesn't really stand out to me any more than proper use of rasterization does... if only we had a 2080 Ti without all the RT and Tensor cores junk that costed cheaper...

 

All this is though based on the fact I want native resolution regardless, I can live without AA as long as it's native rendering, these gimmicks of rendering lower resolution and applying sharpening filters doesn't do it for me... so either ways I needed that extra horsepower.


Workstation Rig:
CPU:  Intel Core i9 9900K @5.0ghz  |~| Cooling: beQuiet! Dark Rock 4 |~|  MOBO: Asus Z390M ROG Maximus XI GENE |~| RAM: 32gb 3333mhz CL15 G.Skill Trident Z RGB |~| GPU: nVidia TITAN V  |~| PSU: beQuiet! Dark Power Pro 11 80Plus Platinum  |~| Boot: Intel 660p 2TB NVMe |~| Storage: 2X4TB HDD 7200rpm Seagate Iron Wolf + 2X2TB SSD SanDisk Ultra |~| Case: Cooler Master Case Pro 3 |~| Display: Acer Predator X34 3440x1440p100hz |~| OS: Windows 10 Pro.
 
Personal Use Rig:
CPU: Intel Core i9 9900 @4.75ghz |~| Cooling: beQuiet! Shadow Rock Slim |~| MOBO: Gigabyte Z390M Gaming mATX|~| RAM: 16gb DDR4 3400mhzCL15 Viper Steel |~| GPU: nVidia Founders Edition RTX 2080 Ti |~| PSU: beQuiet! Straight Power 11 80Plus Gold  |~|  Boot:  Intel 660p 2TB NVMe |~| Storage: 2x2TB SanDisk SSD Ultra 3D |~| Case: Cooler Master Case Pro 3 |~| Display: Viotek GN34CB 3440x1440p100hz |~| OS: Windows 10 Pro.


HTPC / "Console of the house":

CPU: Intel Core i7 8700 @4.45ghz |~| Cooling: Cooler Master Hyper 212X |~| MOBO: Gigabyte Z370M D3H mATX|~| RAM: 16gb DDR4 3333mhzCL16 G.Skill Trident Z |~| GPU: nVidia Founders Edition GTX 1080 Ti |~| PSU: Corsair TX650M 80Plus Gold |~| Boot:  SSD WD Green M.2 2280 240GB |~| Storage: 1x3TB HDD 7200rpm Seagate Barracuda + SanDisk Ultra 3D 1TB |~| Case: Fractal Design Meshify C Mini |~| Display: Toshiba UL7A 4K/60hz |~| OS: Windows 10 Pro.
Link to post
Share on other sites
Posted · Original PosterOP
19 minutes ago, Princess Luna said:

Yeah, I end up going all out on the 2080 Ti at the end of the day, from experience with my 1080 Ti there's the need to fine tune every game to keep on the 100fps margin (also 100hz 1440p ultrawide), considering the 5700XT is ever so slightly slower than the 1080 Ti on average when you OC both I felt that AMD is really out of the picture for me as I wanted to keep as close to ultras as possible.

 

I waited for the 2080 Super but at the end of the day nVidia brought us a card that is 8% faster on average than a 1080 Ti at the same price point, disappointing specially when latest AAA games are even more demanding like Metro Exodus and upcoming RDR2 and Cyberpunk 77... so I just accepted the massive expensive price tag on the 2080 Ti to be safe.

 

Sad part is that I don't give a crap to Ray Tracing, it doesn't really stand out to me any more than proper use of rasterization does... if only we had a 2080 Ti without all the RT and Tensor cores junk that costed cheaper...

 

All this is though based on the fact I want native resolution regardless, I can live without AA as long as it's native rendering, these gimmicks of rendering lower resolution and applying sharpening filters doesn't do it for me... so either ways I needed that extra horsepower.

Yeah I was also awaiting the 2080 Super but that was a great dissapointment so for me it's basically 5700XT, 2070 Super or 2080 non-super that are on the table (Seeing as I game way to few times to justify the 2080TI-pricepoint). 

 

As you say it basically comes down to if you can live with upscaled resolution or not. As long as there is a universal option for the scaling I dont really care if im looking at native or 70-80% upscaled provided that I cannot tell the difference without looking very hard in to the grass or something, so that trade-off turns 5700XT in to a very viable option for me. 

 

F-16.jpg

Link to post
Share on other sites
3 minutes ago, Lindstroem said:

.

70% resolution scaling, no sharpening filter can truly mask that, careful not to be disappointed.


Workstation Rig:
CPU:  Intel Core i9 9900K @5.0ghz  |~| Cooling: beQuiet! Dark Rock 4 |~|  MOBO: Asus Z390M ROG Maximus XI GENE |~| RAM: 32gb 3333mhz CL15 G.Skill Trident Z RGB |~| GPU: nVidia TITAN V  |~| PSU: beQuiet! Dark Power Pro 11 80Plus Platinum  |~| Boot: Intel 660p 2TB NVMe |~| Storage: 2X4TB HDD 7200rpm Seagate Iron Wolf + 2X2TB SSD SanDisk Ultra |~| Case: Cooler Master Case Pro 3 |~| Display: Acer Predator X34 3440x1440p100hz |~| OS: Windows 10 Pro.
 
Personal Use Rig:
CPU: Intel Core i9 9900 @4.75ghz |~| Cooling: beQuiet! Shadow Rock Slim |~| MOBO: Gigabyte Z390M Gaming mATX|~| RAM: 16gb DDR4 3400mhzCL15 Viper Steel |~| GPU: nVidia Founders Edition RTX 2080 Ti |~| PSU: beQuiet! Straight Power 11 80Plus Gold  |~|  Boot:  Intel 660p 2TB NVMe |~| Storage: 2x2TB SanDisk SSD Ultra 3D |~| Case: Cooler Master Case Pro 3 |~| Display: Viotek GN34CB 3440x1440p100hz |~| OS: Windows 10 Pro.


HTPC / "Console of the house":

CPU: Intel Core i7 8700 @4.45ghz |~| Cooling: Cooler Master Hyper 212X |~| MOBO: Gigabyte Z370M D3H mATX|~| RAM: 16gb DDR4 3333mhzCL16 G.Skill Trident Z |~| GPU: nVidia Founders Edition GTX 1080 Ti |~| PSU: Corsair TX650M 80Plus Gold |~| Boot:  SSD WD Green M.2 2280 240GB |~| Storage: 1x3TB HDD 7200rpm Seagate Barracuda + SanDisk Ultra 3D 1TB |~| Case: Fractal Design Meshify C Mini |~| Display: Toshiba UL7A 4K/60hz |~| OS: Windows 10 Pro.
Link to post
Share on other sites

For 1440p uw the 2080 ti is perfect if you can afford it (did the math with the new screens), that said the 5700XT is the best bang for buck in that range. The upscaling is indeed impressive, but i'd wait if i can atm since the consoles next year, which is based on navi will have ray tracing, and no doubt this kind of upscaling will be used, I'm hyped for next year lol.


9900k 1.32v max 1.26v avx 5.1ghz-1avx 4.8 cache 95C 175w 1.05v 4.4ghz 95w 55C R20/blender temps ll D15 ll Z390 taichi ult 1.60 bios ll gskill 2x8gb 16-16-16-34-280-24 ddr3866  bdie 1.42v dram 1.22v io/sa (anything higher needs more voltage on all (dram/io/sa) ll EVGA 2080 ti XC 1995//7600 power limited 79C max, stock voltage (bad ocer) ll 2x samsung 860 evo 500gb raid 0 ll 500gb nvme 970 evo ll Corsair graphite 780T ll EVGA G2 1300w ll Windows 10 Pro ll NEC PA272w (movie, work mon) 2k60 14bit lut ll Predator X27 4k144 hdr (using at 4:4:4 98)

 

old rig 8600k d15 evga 1080 ti hybrid  z370 extreme 4 2x8gb ddr3000 512gb nvme evo+ 860 evo 500gb raid 0 evga p2 1000w 

Link to post
Share on other sites
Posted · Original PosterOP
3 minutes ago, Princess Luna said:

70% resolution scaling, no sharpening filter can truly mask that, careful not to be disappointed.

Not being overly picky with visuals, the experience statements from Hardware Unboxed do inspire me with some confidence, atleast sufficient for me to try to stay my hand from the buy-button on a 2070 Super until the AIB 5700 XT cards comes out. Provided that the scaling works as well as it does (according to the tests made) it is a far more interesting feature for me than Ray tracing (at this time). But will try to test it before a purchase, or hope for more reviews to be made.

"We turned on sharpening for the 1800p image and… wow. We were seriously impressed with the results here. It’s not a perfect recreation of 4K and still suffers from some of the general issues with post process sharpening, but in many scenes it’s incredibly close and virtually indistinguishable from the native image. True 4K has finer detail in foliage and other small elements, but you’d be hard pressed to spot a difference, especially without a side by side comparison."

Link to post
Share on other sites
Posted · Original PosterOP
Just now, xg32 said:

For 1440p uw the 2080 ti is perfect if you can afford it (did the math with the new screens), that said the 5700XT is the best bang for buck in that range. The upscaling is indeed impressive, but i'd wait if i can atm since the consoles next year, which is based on navi will have ray tracing, and no doubt this kind of upscaling will be used, I'm hyped for next year lol.

Yeah going for the 5700 XT now, and then if a significantly improved card with 100% RDNA and refined hardware is released 2020/2021, Im thinking I can sell and upgrade with perhaps a 200 USD drop in price without feeling to bad :)

Link to post
Share on other sites
12 minutes ago, Lindstroem said:

Yeah I was also awaiting the 2080 Super but that was a great dissapointment so for me it's basically 5700XT, 2070 Super or 2080 non-super that are on the table (Seeing as I game way to few times to justify the 2080TI-pricepoint). 

 

As you say it basically comes down to if you can live with upscaled resolution or not. As long as there is a universal option for the scaling I dont really care if im looking at native or 70-80% upscaled provided that I cannot tell the difference without looking very hard in to the grass or something, so that trade-off turns 5700XT in to a very viable option for me. 

 

F-16.jpg

This picture can be applied to the Ray Tracing argument as well.  Standard lighting vs Ray Tracing when in motion and focusing on the game... virtually indistinguishable as well.

 

Or can be said for High vs Ultra settings.  Or a myriad of other things when you tweak various settings.

 

If you want to go by what your eyes can tell or you sorta feel...  most cards perform similarly if you go more subjective and less objective.  

 

Also, I really don't like the "math" basis of comparing the cards.  Use a quantitative measurement when comparing.


Emma : i7 8700K @5.0Ghz - Gigabyte AORUS z370 Gaming 5 - ThermalTake Water 3.0 Ultimate - G. Skill Ripjaws V 32GB 3200Mhz - Gigabyte AORUS 1080Ti - 750 EVO 512GB + 2x 860 EVO 1TB M.2 (RAID 0) - EVGA Supernova 650 P2 - Fractal Design Define R6 - AOC AGON 35" 3440x1440 100Hz - Mackie CR5BT - Logitech G910, G502, G933 - Cooler Master Universal Graphics Card Holder

 

Plex : Ryzen 5 1600 (stock) - Gigabyte B450M-DS3H - G. Skill Ripjaws V 8GB 2400Mhz - GeForce 8800GTS 640MB - 840 EVO 256GB + Toshiba P300 3TB - TP-Link AC1900 PCIe Wifi - Cooler Master MasterWatt Lite 500 - Antec Nine Hundred - Dell 19" 4:3

 

Lenovo 720S Touch 15.6" - i7 7700HQ, 16GB RAM, 512GB NVMe SSD, 1050Ti, 4K touchscreen

 

MSI GF62 - i7 7700HQ, 16GB 2400 MHz RAM, 256GB NVMe SSD + 1TB 7200rpm HDD, 1050Ti

Link to post
Share on other sites
Posted · Original PosterOP
11 minutes ago, jstudrawa said:

This picture can be applied to the Ray Tracing argument as well.  Standard lighting vs Ray Tracing when in motion and focusing on the game... virtually indistinguishable as well.

 

Or can be said for High vs Ultra settings.  Or a myriad of other things when you tweak various settings.

 

If you want to go by what your eyes can tell or you sorta feel...  most cards perform similarly if you go more subjective and less objective.  

 

Also, I really don't like the "math" basis of comparing the cards.  Use a quantitative measurement when comparing.

Can't really comment on the indistinguishable arguments as I myself havent compared GPU scaling+RIS with the native 4k other than viewing it on the Youtube-videos and not being able to really spot a difference even when zooming in and freezing frames (especially not if I would know which where which, which I feel is not the case with RT and high/ultra). 

 

 

Me too would like to see quantitative measurements, I have asked techradar to follow up with such a test to gauge the performance in a multitude of titles as well. 

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


×