Jump to content

The RX 480, Whats wrong with the Tflops?

SEPH

Ok so i might be doing smthing rly stupid writing this but i just don't get it. The RX 480 has 5.833 Tflops of compute at stock boost clocks (1266Mhz), the r9 390 has 5.1 Tflops of compute at its stock clock of 1Ghz, lets give it a little boost to 1100Mhz, that takes it to 5.6 Tflops. So what i dont get is how the 390 is outpreforming the RX 480 when the 480 is a card based on a newer itteration of GCN (AMD claims 15% improved IPC, that should easily make up for those 256 cores missing form the 480 even if its more like 10% better IPC) with faster clocks and whatnot. I will also add that the OC to 1350Mhz on the RX 480 takes it to 6.22 Tflops and even that doesn't beat the 390's 5.8 Tflops (with OC). So i am just a bit lost in this, the r9 390X has 5.8 Tflops so what the hell, It's probably drivers or smthing but to me thats a big gap for drivers especially considering the fact that It's still GCN. Do any of you guys have any ideas as to why this is happening? Anyway, waiting for the custom coolers maybe things will become more transparent when those arrive to market. 

Link to comment
Share on other sites

Link to post
Share on other sites

The card is not fully utilized, that's why it's to slow compared to it's raw performance. Just look at the TFlops of the 970 or 1070 and compare that to the FPS in games.

Mineral oil and 40 kg aluminium heat sinks are a perfect combination: 73 cores and a Titan X, Twenty Thousand Leagues Under the Oil

Link to comment
Share on other sites

Link to post
Share on other sites

You can't use TFlops to determine performance. That's evident when you compare the Fury X with the GTX 980Ti.

'Fanboyism is stupid' - someone on this forum.

Be nice to each other boys and girls. And don't cheap out on a power supply.

Spoiler

CPU: Intel Core i7 4790K - 4.5 GHz | Motherboard: ASUS MAXIMUS VII HERO | RAM: 32GB Corsair Vengeance Pro DDR3 | SSD: Samsung 850 EVO - 500GB | GPU: MSI GTX 980 Ti Gaming 6GB | PSU: EVGA SuperNOVA 650 G2 | Case: NZXT Phantom 530 | Cooling: CRYORIG R1 Ultimate | Monitor: ASUS ROG Swift PG279Q | Peripherals: Corsair Vengeance K70 and Razer DeathAdder

 

Link to comment
Share on other sites

Link to post
Share on other sites

10 minutes ago, HKZeroFive said:

You can't use TFlops to determine performance. That's evident when you compare the Fury X with the GTX 980Ti.

i know but the thing is this isn't AMD vs. Nvidia its AMD vs. AMD and GCN vs. GCN albeit different itterations and the newer version with the higher Tflops is losing??? like wtf

Link to comment
Share on other sites

Link to post
Share on other sites

26 minutes ago, SEPH said:

i know but the thing is this isn't AMD vs. Nvidia its AMD vs. AMD and GCN vs. GCN albeit different itterations and the newer version with the higher Tflops is losing??? like wtf

It's a very common misconception.

 

TFlops isn't a measure of gaming performance. Made even more so by the number of other variables. It's a measure of peak compute power.

 

Think of it as comparing the clockspeeds of different CPUs. It's utterly useless in determining relative performance.

'Fanboyism is stupid' - someone on this forum.

Be nice to each other boys and girls. And don't cheap out on a power supply.

Spoiler

CPU: Intel Core i7 4790K - 4.5 GHz | Motherboard: ASUS MAXIMUS VII HERO | RAM: 32GB Corsair Vengeance Pro DDR3 | SSD: Samsung 850 EVO - 500GB | GPU: MSI GTX 980 Ti Gaming 6GB | PSU: EVGA SuperNOVA 650 G2 | Case: NZXT Phantom 530 | Cooling: CRYORIG R1 Ultimate | Monitor: ASUS ROG Swift PG279Q | Peripherals: Corsair Vengeance K70 and Razer DeathAdder

 

Link to comment
Share on other sites

Link to post
Share on other sites

teraflops if the theoretical peak compute performance, drivers have not matured yet, more performance is to be expected from that card with newer drivers

as evidenced by the promisse to considerably boost GTA V performance on RX 480 and all GCN 1.1 and up cards with upcomming updates

 

theoretical peak compute performance means that with perfect software that is the absolute maximum number of floating point operations you can achieve per second, it's a good measure to evaluate the efficiency of software, it's mostly pretty easy to get around 50~70% of the performance you can push it to 90% with very clever software, going above that is insanely expensive and doesn't really pay off for the effort, 100% theoretical performance is very rarely if ever achieved

CPU: Intel i7 5820K @ 4.20 GHz | MotherboardMSI X99S SLI PLUS | RAM: Corsair LPX 16GB DDR4 @ 2666MHz | GPU: Sapphire R9 Fury (x2 CrossFire)
Storage: Samsung 950Pro 512GB // OCZ Vector150 240GB // Seagate 1TB | PSU: Seasonic 1050 Snow Silent | Case: NZXT H440 | Cooling: Nepton 240M
FireStrike // Extreme // Ultra // 8K // 16K

 

Link to comment
Share on other sites

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

×