Comments on: AMD R9 280X Crossfire Results & Benchmarks!

By: JumboJelly

JumboJelly — Sun, 27 Oct 2013 04:23:00 +0000

Elric,

I’m thinking of purchasing the Asus r9 280x, but I have no benchmarks to finalize my decision! Please make some benchmarks of a single Asus 280x against some other cars like the 770 from nvidia.

By: Serpent of Darkness

Serpent of Darkness — Wed, 23 Oct 2013 00:06:00 +0000

The HIS 7970×2 is a dual GPU card. It’s two GPU in CrossfireX with one another, on a single PCB. If you compare them to two single 7970s in CrossfireX, the FPS correlation scales a little bit better. The idea behind the benchmarks is to give you some kind of foundation, or a comparison between dual GPU PCB-cards, a single, and two singles in CrossfireX. In addition, Elric has a GTX 770 and 780 card to compare the performance with the CrossfireX 7970 and RX9-280 cards. The Graph indicates that the RX9-280 and 7970 scale a lot better, in CrossfireX than the dual GPU PCB-cards. Scale by a factor of 1.90013 at 1080p, and the scaling is at 1.9205 at 1600p. So one person can come to the conclusion that Crossfire actually scales a little bit better at higher resolutions by 10.72%. Sadly, there should have been an SLI comparison, but you can probably crunch the numbers. Lets just focus on BF3. In 1080p, there’s a 90% increase in FPS performance with the CrossfireX setup versus a single RX9-280, and there is a 92.1% increase on 1600p. Let’s say SLI setups scale the same way. So GTX 770 will have 152.83 fps on 1080p, and it will have 94.4 fps on 1600p. GTX 780 will show FPS of 173.9 fps on 1080p, and in 1600p, the fps performance will be 110.43 fps. What’s the conclusion, CrossfireX scales rather well with the new RX9-280, and the performance is somewhere in between GTX 770 in SLI and GTX 780 in SLI–that’s if you do the math, or compare it to other 3rd party benchmarks.

By: Diego Braz

Diego Braz — Tue, 22 Oct 2013 17:03:00 +0000

whats the power supply???

By: Mlgpro Marik

Mlgpro Marik — Mon, 21 Oct 2013 20:42:00 +0000

its not about the money, they are comparing 2 cards vs 1…

By: g0ggy

g0ggy — Mon, 21 Oct 2013 20:28:00 +0000

Actually some people would, especially in countries where electricity is extremely expensive such as Germany.

By: Andrejs Silvans

Andrejs Silvans — Mon, 21 Oct 2013 19:39:00 +0000

Someone that buys a card for $1000 does not care about his electricity bill. Sorry still gotta go with the 280x xfire solution.

By: Serpent of Darkness

Serpent of Darkness — Mon, 21 Oct 2013 06:42:00 +0000

In the last paragraph, I am referring too PCIe CrossfireX that will be used on the RX9-290. A lot of people feel it’s not a big deal, but it is.

By: Serpent of Darkness

Serpent of Darkness — Mon, 21 Oct 2013 06:30:00 +0000

Nice benches. Wish Elrich did his own graphs on Frame Time Latency and his own explanation of what they mean, but not getting to nerdy about it. I feel he would do a great job communicating this to his fan base.
Saw the video from Tek Syndicate. Some of the graphs like the Trine Graphs, has a larger Frame Time Band. I think it’s due to the fact that the RX9-280 and 7970 GHz, there’s a latency or delay between each gpu when they fire frames from the Fame Buffer. If I am not mistaken, CrossfireX Renders frames from each Graphic card. Primary goes first for Frame 1. Then the buffer card goes second for frame 2, and it’s a repeating process. So what essentially happens, in Trine, with CrossfireX is that one Frame Buffer will fire at a really small Frame Time. This translate to a higher FPS. 13 ms? It could be smaller than 13 ms. Then it jumps up to like 25 ms, then the 2nd Frame Buffers fire a frame to the monitor, and the graph spikes down to like 14 or 13 ms again, and so on. The process just repeats itself. This is good and bad. It’s good because the lower the frame time, the higher the FPS. In between each graphic card, it’s producing high FPS, or 1st derivatives. It’s bad because there’s a window of time in between each GPU firing a frame from their frame buffers. Remember, the GPU and the CPU communicate, the GPU basically creates the frame, and stores this information to the frame buffer. Following after that, it gets sent to the Monitor to be displayed. A simplified idea of what I am saying is like shooting two hand guns off, but not at the same time. You shoot one off, then you discharge the other gun. Think of the bullet coming out of the gun as a Frame being sent to the monitor. Lag between the two GPUs sending that frame, is the time it takes in between gun 1 and gun 2 discharging.
I think it’s ignorant of NVidia to reveal G-Sync. I’m not trying to go off topic, but basically, if you read the nitty gritty, it’s basically saying that the monitor is actually the issue with Tears (two separate, but partial frames being displayed), ghosting (some delay in the transition of frames), micro-stutters(fames not being displayed because they have a higher frame-time aka lower fps), and runt-frames (partial frames from two sources (cards) being displayed). You’re monitor’s static refresh rate is really the issue. That 60 Hz refresh rate is what I am referring too. Simple and short, if frames aren’t produced under 16.667 ms, you’ll run into problems. This generation of graphic cards are so OP that the displays are lacking. So NVidia solved the problem with G-Sync.
I think with PCIe through the PCI Bus, this could solve some problems with CrossfireX. The delay time in between firing off Frame Buffers 1, 2, and 3–looking at three graphic cards in CrossfireX, will get smaller. So this will shrink the deviation band in AMD graphic cards, when you look at Frame Time Latency Graphs. Also, through the PCIe Bus, they could probably better control when Frame Buffer 1, 2, 3, and 4 can fire off frames to the display, within that 16.667 ms through software, and it wouldn’t be necessary for it to occur on a hardware level.
Another issue I can see with CrossfireX, and this is probably addressed with AMD Beta Drivers with Frame Pacing Software on them, is the fact that GPU 1, 2, 3, and 4 will attempt to ram 3 to 4 frames in that 16.667 ms window. That’s assuming they aren’t dropped because newer frames are produced. I suspect, without the beta driver, there’s nothing that actually regulates this. AMD Graphic Cards are like Race Horses. They want to ram frames into this 16.667 ms window of time, to display images, as fast as possible. When they can do it under the 16.667 ms window, I suspect the display takes on parts of the 2nd frame, from the 2nd frame buffer. Thus, a runt frame is produced…

By: Ze Übermensch

Ze Übermensch — Sun, 20 Oct 2013 18:31:00 +0000

Performs pretty well.
#givemethecard

By: luke

luke — Sun, 20 Oct 2013 14:28:00 +0000

well then let them get the one wich is more expensive if money is not an issue , 2 280x cost as 1 780 sooo