dj-electricAMD needs to stop giving new versions of its software new names.

This is ridicules. Some small, and respectable changes, and that's it.

dj-electricAMD needs to stop giving new versions of its software new names.

This is ridicules. Some small, and respectable changes, and that's it.

AldainThis is not a new software version name . it is a name of the project. How about you read first then comment

cucker tarlsonHow about working on that friggin dx11 cpu overhead first ?

krukHow about developers doing some friggin optimizations or switch to another, better API?

There is simply no excuse for sh*tty programming, because we have several visually stunning games (DOOM, BF1, etc.) that run excellent on both AMD and nVidia ...

HD64GI cannot get why some people here bash AMD for making gaming better on a game genre so popular.

bugBetter? All they have to show for their effors is 11% in one game, in one specific configuration. I don't know where you've been living, but in real life this is called... wait for it... a regular driver update.

HD64GI cannot get why some people here bash AMD for making gaming better on a game genre so popular. Some do the same just because AMD GPUs are better in compute performance and they are preferred from miners. But AMD GPUs are best in compute power due to CGN arch since the start (7970) before mining became a thing. I guess haters gonna hate... :shadedshu:

renz496if AMD card really are the best in compute then their Firepro will be used more in professional space. AMD GPU was better at mining not because they have better compute performance than nvidia but more mining algo was built to take advantage of AMD architecture. but the same can also be done on nvidia hardware. i don't know why some people look at AMD superior performance at mining (and mining only) and then assume AMD as better at compute compared to nvidia. in reality while AMD GPU theoretically have much higher compute peak performance than nvidia most often it is hard for them to properly utilize all that raw power compared to nvidia. there was two super computer that rank almost next to each other at TOP500 list. both are using accelerator. the one that using AMD accelerator (tahiti based) have much higher peak performance than the one using nvidia tesla. but nvidia based machine end up being much better rank because it can utilize it's raw power better and having much faster real world performance.

HD64GIncrease in performance in any game is considered better than no increase, even if it is less than 10%. Where is the bad thing here that deserves critisism?

HD64GYou are clearly confusing hw and sw. AMD is usually better in raw compute power due to its arch and nVidia is better in specific tasks due to their focus on sw. 1080Ti is the exception.

renz496And they build their hardware in a way that it is more difficult for the software to utilize the raw performance. Just look at async compute for example. They need specific support from the API before they can use them. Nvidia primary improvement on maxwell (over kepler) also increasing GPU utilization but they do it in a more transparent to the game hence regardless of API being use by the game they can benefit from the improvement done on maxwell.

Also most of the complaint here is not so much about the minimal increase performance but more on AMD keep getting out with fancy names with everything they do. Nvidia also something similar before (reducing latency on esports game) when they launch GTX950 but they never did give any fancy names for it. It is just standard driver improvement.

renz496And they build their hardware in a way that it is more difficult for the software to utilize the raw performance. Just look at async compute for example. They need specific support from the API before they can use them. Nvidia primary improvement on maxwell (over kepler) also increasing GPU utilization but they do it in a more transparent to the game hence regardless of API being use by the game they can benefit from the improvement done on maxwell.

Also most of the complaint here is not so much about the minimal increase performance but more on AMD keep getting out with fancy names with everything they do. Nvidia also something similar before (reducing latency on esports game) when they launch GTX950 but they never did give any fancy names for it. It is just standard driver improvement.

theoneandonlymrkWhat fancy name 18.3.1??

This is an internal work effort code name , even in pr blurb it is not a new driver name calm down.

Also it's Nvidia that has the preferance for special coding ala cuda and some scenarios favour Nvidia because Nvidia work very hard on the software

And people call Nvidia out on compute because they culled some of it totally out the low to mid range and also now charge a premium for it's use or in some cases more like a triple premium ala quadro.

bugTbh what you're describing is an old problem. Intel had the same problem with Itanium. Programming a RISC CPU used to be way harder then programming a CISC CPU.
AMD having the same problem doesn't mean they don't know how to build their hardware. It's just that things don't always work out in practice as well as they do on paper. And on top of that, you're supposed to be designing for years ahead.
Still, their driver overhead problem is old and they should have addressed it by now, imho.

renz496"Project " is the fancy name. Regarding to compute honestly i find it really funny though. When nvidia include lots of compute stuff with fermi people say nvidia selling useless feature to gamer. Then when AMD follow suit and nvidia start seperating compute feature from consumer card suddenly compute become important for a gaming card.



Not saying that AMD did not know to make their hardware but they might need to rethink the way they built their hardware. Just because now there is DX12 then DX11 is useless. Just look at Vega. When the game was specifically built to take it's feature it can be very fast. But reality is gamer did not play new games only. New feature is nice but they also need to think about existing software.

bugAMD's drivers have historically been more CPU intensive than Nvidia's when it comes to DX (and OpenGL). How can a developer work around that? What's the job of a driver programmer if not squeezing everything that can be squeezed from the hardware?