no 20nm gpu for 2015. Probably ever. They may skip to different node.

New leaked suggested the FIJI will be 300W TDP. If its true, it's so obvious it will be 28nm. The TDP is just

It's an issues ever since LCD was invented. People are so use to the 60hz/fixed refresh rate, they don't even thought it need to be fixed. Until nvidia came up with g-sync, the industries just so comfortable with fixed refresh rate.

Lot of things can be solved with VRR because media sources are recorded/film with different frame rate at different refresh rate. So this need to be address. We don't want to have fixed refresh rate, capped fps for games and duplicated frames and odd frames for movies that run at 25fps at 60hz.

VRR on monitor is the first step to fix all this issues. Hopefully this will come to consumer electronics (LCD tv, stb and what not).

I'm looking forward they fixed this first before going to 4k/5k/8k.

This post has been edited by ruffstuff: Mar 11 2015, 04:56 PM

below 30fps is not tearing artifact, it is called stuttering.

when the graphics card push above your refresh rate, than it is tearing.

This post has been edited by ruffstuff: Mar 20 2015, 03:52 PM

It is very hard to explain VRR, or to show it on videos etc. You have to experience it yourself, then you know the difference.

G-sync and freesync tested via oscilloscope. It is important how both technology handle VRR below the panel refresh rate.

It looks like G-sync deal with low fps smarter than freesync. In a nutshell, freesync just stop doing it works at below panel refresh rate, leave you with the classic stuttering and tearing. Meanwhile, g-sync doing frames interpolation (double or tripple the frames), and match it exactly with the refresh rate. Genius.

This post has been edited by ruffstuff: Mar 28 2015, 03:03 PM

If you already using nvidia, changing to ATI is even costlier.

Im not saying you need to change to nvidia for g-sync.

Another drawbacks of freesync, due to narrow VRR window for this particular monitor.



Although AMD claim the default refresh rate for adaptive sync is from 9-240hz, but this not the case because it goes back to the panel limitation itself.

Although this problem is not AMD freesync specific, it more to the panel itself.

Reading through the VESA white paper for adaptive sync, it seems that the standard do have that doubling/trippling frames feature if the fps drops below VRR window. But, to implement this, the panel itself need frame buffer. This is the missing piece in Freesync technology. The scaler manufacture mostly will not include frame buffer into their scaler for the cost reason. If they do, the price margin between freesync and g-sync is getting narrow.

The g-sync module as we know, do have frame buffer/memory chips.

The refresh rate window is panel specific. The adaptive-sync spec is from 9-240hz, but no way panel manufacturer will allow the panel to refresh as low as 10hz.


At this moment theres no freesync monitor that can do VRR below 30hz that im aware of.

Freesync monitor have refresh rate window from 40hz. Within the Variable refresh rate (VRR) window, the freesync works flawlessly. But thing get really different between freesync and g-sync when the frame rate drop below the vrr window. When fps drop below 40, you'll get the classic stuttered like any other non vrr monitor. In g-sync, they implement the frame interpolation (frame duplicating/doubling/tripling), if the FPS drop below VRR window. For example, if the fps is at 30, g-sync monitor will refresh the panel at 60hz, while it g-sync module buffer the fps by doubling it to match the refresh rate. If the fps at 20, the panel will refresh at 60hz, the g-sync buffer will triple the fps. The g-sync module will keep the panel to refresh at the VRR window, at any fps without odd frames that mismatch the refresh rate.

It is still unknown why freesync unable to do frame doubling, the fact that VESA adaptive standard DP1.2a+ white paper have frame doubling technology too if the fps drop below vrr window. My speculation is, the lack of asics/frame buffer in panel scaler to achieve what g-sync can achieve. I do think this can be done in the GPU buffer itself. Maybe lack of mature drivers?

However, if fps shoot above VRR window, freesync have the option to have v-sync on or v-sync off. While nvidia g-sync will force v-sync on (cap fps) at maximum panel refresh rate.

List of freesync panel and its VRR window.



This post has been edited by ruffstuff: Apr 18 2015, 04:37 PM

The samsung UE590 final spec is not revealed yet. But i bet it is minimum 40hz maximum 60hz.

The 9-240hz is the universal refresh rate space. Not the panel refresh rate. Freesync works within panel refresh rate. The gpu negotiate with panel the VRR window before hand, and AMD claim this is more superior, and have 0 latency than nvidia g-sync.

There is no panel that can run at 9-240hz at this moment.

Should be no problem if framerate within the vrr window. But i would like to see minimum 30hz for 4k panel. Not really concern on max refresh rate because current gen card hardly can output more than 60. Minimum refresh rate is important because you'll encounter frame drops more often on 4k resolution.

Hopefully, but freesync doesn't work on crossfire. Yet.

I wonder what is AMD new mainstream and high performance line up, if FIJI is enthusiast rather than affordable performance card. How they can scale down FIJI chip? The new Tonga chip is not a performance chip, can they make it a new performance chip based on Tonga?

Or AMD user will have to stick with Hawaii rebadge for mainstream performance card?

P/S: I have a strong feeling that FIJI will not be using a R9 390x name scheme. It can be R10 XXX.

This post has been edited by ruffstuff: May 21 2015, 12:05 PM

And it is not R9 390X. It will be call Radeon Fury.

Nope. The maxwell full fat core was core was completed ever sknce the introduction of 980. Before 980 released, people expecting GM200 core to be the 980. But no, it is a crippldf version of GM200. Nvidia dont feel the need to release the full.maxwell core at launch because it isnt necessary. The stupid move of nvidia is the branding of GM200 as the Titan X which is kinda confusing this time around. The Titan X is missing important feateru that distinguish itself as gaming centric card. It doesnt have the fp64 and double precision. Which is strongly suggest this chip is the supposed to be 980 at launch. Well at least they made Ti out of it.

The downside of the card is the chip itself. HBM is great, but what is the point of having small footprint card, but ended up with huge radioator alongside with it? If AMD can make a high performance card with that footprint, without the need of wc, its a clear winner. Even if the performance is slightly below 980. I hope the non XT Fury will be like this. No need wc, but still maintaining the small footprint.

If possible, slightly better performance than 290x, without the watercooling, and maintaining the footprint. Still a good deal.