Yay! I actually finished doing the tests that I said I would do in this post. Whenever I sit down to make a post, I always don't want to type a whole lot, because then I have to get up and walk it off for a while and I give myself a headache. But whatever, I will now start typing this post. >_>
The Cards are on the Table
Puns! But in this case they really are on the table. In any case, I (as I often do) made some adjustments to my test. I didn't test the R9 290X because I wanted to get the 770 and 280X tested and written up first. But I successfully modded my R9 290X with a Kraken G12 and Asetek 570LC AIO cooler, and managed to overclock it to 1200 MHz which is a nice result. I might make a post on that later.
Anyway... I love graphics cards so let's take a look at the cards.
Sapphire Radeon R9 280X Vapor-X
The card on the left is a Sapphire R9 280X Vapor-X. This card launched for about $350-360 making it actually quite expensive for a 280X, but it is a great model. It's still cheaper than the $400 GTX 770, although the price of these did come down. Some basic Tech Specs! Yay!
For those that don't know, the R9 280X is a re-branded HD 7970 GHz Edition which is a Re-branded HD 7970, with higher core clock speeds and faster GDDR5 memory.
Graphics Processor (Silicon)
GPU: "Tahiti XTL"
GPU Family: "Southern Islands"
Silicon Process: TSMC 28nm
Die Size: ~350mm 2
Transistors: ~4.3 billion
Graphics Processor (Specification)
Graphics Architecture: Graphics Core Next 1st Generation (GCN1)
Graphics Cores: 32
Total Stream Processors: 2048
Total Texture Units: 128
Geometry/Tessellation Processors: 2
Raster Engines: 2
L2 Cache: 768 KB
Pixel Pipelines: 32
GPU Engine Clock: 1070 MHz
FP32 Shading Power (Theoretical): 4.3 TFLOPS (FMA)
Memory Subsystem
Memory Type: GDDR5
Memory Speed (Effective): 1550 MHz (6200 MHz)
Memory Capacity: 3 GB
Memory Bus Width: 384-bit
Memory Bandwidth: 297 GB/s
I enjoyed typing that. Anyway, for the GeForce
MSI GeForce GTX 770 Twin Frozr
The card in the middle of the top picture is MSI's Twin Frozr GTX 770. It's a nice cooler but it doesn't come with a back plate which doesn't even make me a Sad Panda because I don't really care. Anyway this card was originally launched for $400, but I think it got a price cut.
For those that don't know, the GTX 770 is a re-branded GTX 680 with faster GDDR5 and a slight increase in core clock speeds.
Graphics Processor (Silicon)
GPU: "GK104-425"
GPU Family: "Kepler (GK10x)"
Silicon Process: TSMC 28nm
Die Size: ~294mm 2
Transistors: ~3.5 billion
Graphics Processor (Specification)
Graphics Architecture: Kepler
Graphics Cores: 8
Total Stream Processors: 1536
Total Texture Units: 128
Geometry/Tessellation Processors: 8
Raster Engines: 4
L2 Cache: 512 KB
Pixel Pipelines: 32
GPU Engine Clock: 1215 MHz (Includes GPU Boost on this specific card).
FP32 Shading Power (Theoretical): 3.7 TFLOPS (FMA)
Memory Subsystem
Memory Type: GDDR5
Memory Speed (Effective): 1750 MHz (7000 MHz)
Memory Capacity: 2 GB
Memory Bus Width: 256-bit
Memory Bandwidth: 224 GB/s
Okay, now that I have got that out of the way, I will type about my test set up, re-iterated from my previous post, and anything else I think is important to disclose.
Test Setup and notes
AMD Ryzen 5 3400G @ stock (4.1 GHz in games) (4 core, 8 thread, 4MB L3)
16GB 3200 MHz 16-18-18 (2x8)
MSI B450M Mortar
WD Green 240 GB M.2 SATA
Sabrent Rocket 256 GB M.2 NVME
Silverstone Strider 1000W PSU
Windows 10 Pro (1903)
AMD Radeon Driver 19.11.1
NVIDIA Driver 441.12
Both cards run at PCI-e 3.0 8x
Both cards do not throttle. This is not a test of thermals.
R9 280X runs 1070 MHz / 6200 MHz
GTX 770 runs 1215 MHz / 7000 MHz
I am fully aware that the Ryzen 5 3400G isn't the Best Gaming Processor In The World, but all tests are GPU bound and results are gathered from at least 3 runs to rule out issues with inconsistency. Anything else is outside of my control (Driver/OS/Game issues for example).
If people that are angry, fanboys and/or hate me, take issue with my test setup; the games I tested; their settings; or anything else I would like to direct their attention to this video which accurately represents my attitude regarding that issue.
I am having a depressive episode but I am going to try and 'power through it'. So I might not be as quirky or say stupid things as much. Or I might do, because I am just like that.
I combined all of the results into one graph, to make it a bit more compact. I will list the settings for each game first.
Settings Used
Metro: Last Light Redux
1920x1080
Quality: Very High
SSAA: Off
Texture Filtering: AF 16X
Motion Blur: Normal
Tessellation: Very High
Advanced PhysX: Off
Batman: Arkham Knight
1920x1080
Texture Resolution: Low
Shadow Quality: High
Level of Detail: High
Motion Blur: On
Chromatic Aberration: On
Film Grain: On
Texture Filtering: 16X Anisotropic
GameWorks: ALL DISABLED
Tomb Raider
Exclusive Full Screen
1920x1080
Quality: Ultra Preset
Motion Blur: Yes
Screen Effects: Yes
Ashes of the Singularity: Escalation
1920x1080
DX12
Vsync/mGPU/Overlays: Off
MSAA: off
Point Light Quality: High
Glare Quality: High Terrain
Object Quality: High
Shading Samples: High
Terrain Shading Samples: High
Shadow Quality: High
Texture Quality: Low
Enviromental Effects: Yes
Cosmetic Clouds: Yes
Far Cry: New Dawn
Ultra Preset
75 FoV
1920x1080
Metro: Exodus
DX12
1920x1080
Quality: Low
Texture Filtering: 16X
Motion Blur: Low
Tessellation: Full
Advanced PhysX: Off
Hairworks: On
Ray Trace: Off (Obviously)
DLSS: Off (Ditto)
Shading Rate: 100%
Gears 5
Medium Preset
1920x1080
Ambient Occlusion Intensity: 11
Particle Spawn Rate: 7
FoV: 90
Results
Summary
The R9 280X Vapor-X is on average across all 7 games I tested, 13% faster. The biggest advantage the Tahiti-based card has is in Metro Exodus, where it is a staggering 100% (twice) faster. This result is interesting as Hairworks is enabled, as it is my preference when playing this game. Despite this huge performance advantage, neither card could really provide a fluid experience in this game.
The R9 280X also has a huge advantage in Far Cry New Dawn, in which (As you can see in the gameplay videos) was the dfifference between an unplayable and playable experience, with the Tahiti-based card delivering a solid, 50FPS~ average. The GTX 770 in the gameplay video often dipped into the sub-30 range, which really hurts fluidity and I didn't think it was a playable experience on the settings used. That is a victory for the R9 280X
Interestingly, the mid-sized Kepler-based card has a slight advantage in Gears 5 and Ashes of the Singularity, of a bout 10%; both of which are DX12, and the former is the newest game that I tested. Unfortunately, both cards were having issues with crashing (display driver in the case of the Radeon) in Gears 5 so I didn't manage to get a gameplay video of the R9 280X due to said crashing. The GTX 770 managed to complete my gameplay scene and I may try and get the 280X to work in the future, but I am too depleted to try now. You may find those videos in the Video section of this post, further down.
As you will also notice in the GTX 770's Gears 5 video; it could be considered something of a pyrrhic victory; as the card's 2GB video memory capacity is having a significant adverse effect on the texture quality in the game. With Gears 5, it seems the game will lower the resolution of rendered textures to fit within the Video Memory, instead of actively seeking to use the system RAM. The result is a smoother experience on low-VRAM cards, but as you can see in the video - the textures on a lot of objects look like a blurry, pixelated smudge.
The R9 280X didn't exhibit the same extent of texture issues during the gameplay I did manage, but it was still occuring on some objects. As I said I will try and get the 280X to actually complete the gameplay without crashing, lol - that in itself is hardly a win for the Radeon. Without issues in other games, I suspect a driver/software issue with the Tahiti card. Potentially something to do with Async Compute, which I know is limited on Tahiti, and the fact that it is enabled in the game for this card. Also it is worth noting that AMD is almost definitely not targeting this GPU with fixes/optimisations today.
Overall, the biggest win is in Far Cry New Dawn. The R9 280X's performance advantage was nearly 50%, and the difference between playable and uncomfortable in the gameplay video.
Metro Exodus wouldn't be fully playable on either card even at low quality settings, though you can see in the Gameplay videos of the "Volga" map, that the 280X did a decent job of keeping it most above 30 FPS, where the GTX 770 has significantly worse performance. However, as the benchmark shows, when you reach the more demanding "Taiga" map, neither card can keep the game playable at all. So this is not really a true win for Tahiti in my eyes - only a synthetic one.
The other games were too close for me to call them anything more than a "synthetic victory" for the victorious card.
Gameplay Videos
Here you can find the gameplay videos for Gears 5, Metro Exodus and Far Cry New Dawn. Only the GTX 770's video is available at this time for Gears 5, until I can figure out what was causing the display driver resets.
Metro Exodus on R9 280X Vapor-X
Metro Exodus on GTX 770 Twin Frozr
Far Cry New Dawn on R9 280X Vapor-X
Far Cry New Dawn on GTX 770 Twin Frozr
Gears 5 on R9 280X Vapor-X
(Potentially coming soon if I can fix the driver crashing)
Gears 5 on GTX 770 Twin Frozr
(if these haven't loaded yet, or at low quality, please check back later when they have finished processing).
Conclusion
The R9 280X Vapor-X is a faster video card than the GTX 770 Twin Frozr, in the games I tested, today in 2019. It also has 1GB more video memory, which can make a big difference. All of the games I tested, except for Metro Last Light Redux, used all of the 2GB on the 770. Far Cry even maxed out the 3GB on the 280X. Video memory alone adds to the longevity of a graphics card and the Kepler generation of cards learned that the hard way.
The R9 280X's biggest win is Far Cry New Dawn, where it is able to provide a reasonable 1080p gaming experience on Ultra Settings, something the GTX 770 cannot do. While the R9 280X is immensely faster in Metro Exodus, you would need to upgrade either card to play this game at any capacity at 1080p. The Kepler-based card has very little going for it, as the only game it wins without a caveat is Ashes of the Singularity and only by about 10%. This is a surprising result and has me wondering if it is a limitation of the Tahiti chip, as its advanced compute capabilities in DX12 have been brought into question (re: Async isn't fully used).
While the GTX 770 does provide a higher average frame rate in the latest Gears of War Game, Gears 5, it does so at the expense of very poor image quality. Something the 3GB memory on the R9 280X would no doubt help with, even at the cost of slightly lower averages.
Overall, the R9 280X is a superior graphics card in my test, and underscores the advantages of having more video memory, especially back in 2013. It also demonstrates that AMD's GCN design has aged a bit better than Kepler, especially as developers over the years after Tahiti's launch, have been able to at least attempt to optimsied their codepaths for GCN since it was in the popular consoles.
Anyone who bought an HD 7970, 7 years ago, still has servicable gaming performance today. At least, in the games I tested. I can't say that for the GTX 770 2GB - but it would have been interesting to test the 4GB version of that card, as it would have addressed many of its issues, especially in Gears 5, where if not for the texture issues, the 770 would have been the superior card.
Thanks for reading.