• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

****Official OcUK Fury X Review Thread****

Not really - it is only because of HARDOCP's stupid way of doing benches. The 290x is using lower graphics settings on almost all the games.

Ah ok. edit...

Its not 4K anyway. i'm pretty sure he said 4K
 
Last edited:
I would like to see it in Mantle and DX12 or higher 1440P IQ settings, i have a feeling it may be Nvidia's better Driver Overheads.

AMD's Driver is struggling to feed the card very high Draw Calls and high frame rates.

Well to be honest that's a good point as the ace units and global data share in tonga is mostly unused when it comes to game engines, (it's to wide for current game engines bar a few). Direct x 12 claims to improve the coding for more parallel instructions so maybe you are right. But as it stands I still believe tongas front end is too inefficient and unutilised for the current games so far. (It's the only reason i can think of the irregular performance, because it's similar to how tonga and tahiti look when compared in games.
 
Ofc it's about driver overhead. Even 290x has a lot of problems getting fed fully on 1080p. And in games like Project Cars it really is showing.

He're a good video that really shows how bad AMD overhead is:

https://www.youtube.com/watch?v=IpATnpx45BI

Just take a loog at gpu usages, and that's on 280x.
60-65% gpu usage on win10 1023.0 drivers

So just think how low gpu usage fury-x will have. I bet it has like 30% gpu usage under those settings. Card is basicly idling.
 
Do you think Win10 will improve matters? Even in DX11 games?

It does because of WDDM 2.0. But it won't be enough as it is. If you look at the video I linked, it did improve usage, but still nowhere near 100%. AMD really needs to work on their DX11 overhead.

DX12 ofc won't have this problem as there's plenty of cpu usage there to get 100% usage, but DX12 games will take time to appear.
 
Did any of the reviews run Futuremark's API overhead test? Not against other cards, but to compare DX11 with DX12 / Mantle.

Star Swarm would also be a good one.
 
Could explain why the X99 systems appear to be performing better in reviews.

Hmmmm hes pretty much claiming there that the Fury uses aggressive swapping in and out of system RAM when 4GB isn't enough - which means the limit is the 16GB/s PCI-e bus (or worse if your not on a fully functioning PCI-e 3.0 pipeline) and the system memory bandwidth* while trying to put the emphasis on how fast the HBM is which makes no odds when its not the weakest link in the chain.


* Just because you have say DDR3-2400 in dual channel rated at 25GB/s doesn't mean that at anyone one time the system can pump 25GB/s at the GPU - so having DDR4/quad channel, etc. memory setups (i.e. X99) could possibly be an advantage as they'd more often have more bandwidth available to the GPU.
 
Well to be honest that's a good point as the ace units and global data share in tonga is mostly unused when it comes to game engines, (it's to wide for current game engines bar a few). Direct x 12 claims to improve the coding for more parallel instructions so maybe you are right. But as it stands I still believe tongas front end is too inefficient and unutilised for the current games so far. (It's the only reason i can think of the irregular performance, because it's similar to how tonga and tahiti look when compared in games.

Hawaii also has 8 ACE units.

Its probably a bit late now for AMD to worry about their DX11 driver overheads. win 10 is out next month and DX12 with it.

To those who say we ain't going to see DX12 games for another year, personally i think you're quite wrong.

I don't want to sound the perpetual Fiji optimist, i'm still disappointed with it, but i think there might be an explainable reason for the lack of lower res performance, i think Fiji might just come into its own in DX12 / Mantle.
 
Did any of the reviews run Futuremark's API overhead test? Not against other cards, but to compare DX11 with DX12 / Mantle.

Star Swarm would also be a good one.

Star Swarm would have been interesting to see. But overhead test won't tell us much. Those drivers give around 1.2M on W7 , and 1.4M on Win 8.1. (4770k @4Ghz).

And DX 12 and Mantle can max cards command processor if you have 6 cores (Roughly 20-22M). Ofc we don't know if Fury X has better command processor than 290x, though I don't see a need for one as we won't be limited by 20M in a long long time.
 
Hawaii also has 8 ACE units.

Its probably a bit late now for AMD to worry about their DX11 driver overheads. win 10 is out next month and DX12 with it.

To those who say we ain't going to see DX12 games for another year, personally i think you're quite wrong.

I don't want to sound the perpetual Fiji optimist, i'm still disappointed with it, but i think there might be an explainable reason for the lack of lower res performance, i think Fiji might just come into its own in DX12 / Mantle.

Yup the way I'm looking at it is it does trade blows with Ti and TX wins some and loses some on current games.

I.e Fury X is fine for current DX11 games, and soon the transition to DX12 will start.

Things could get better with DX12, newer drivers, custom BIOS and voltage control. Performance is already strong. It's definitely a good effort from AMD imho. Looking forward to testing the card and for me personally will enjoy using the Fury X until we get a die shrink etc next year.

Been on Nvidia for a while now, fancy a switch back to the red team. A few extra FPS either doesn't influence my decision as any of these current cards is arguably good enough for current games. I just fancy something new to play with. Not as if we are going to keep any of these cards forever. Buy, sell rinse & repeat.

Plan is Titan X > Fury X > Nvidia Pascal Titan > ???
 
Hawaii also has 8 ACE units.

Its probably a bit late now for AMD to worry about their DX11 driver overheads. win 10 is out next month and DX12 with it.

To those who say we ain't going to see DX12 games for another year, personally i think you're quite wrong.

I don't want to sound the perpetual Fiji optimist, i'm still disappointed with it, but i think there might be an explainable reason for the lack of lower res performance, i think Fiji might just come into its own in DX12 / Mantle.

I share your optimism for Fiji performing well under DX12. But I don't share your optimism when it comes to DX12 games. Surely there will be games out with DX12, but majority of games will ship with DX11 only. Especially games made by smaller studioes who haven't had change to get in DX12 early access programs. Though I hope I'm wrong and we see atleast 50% DX12 adaption in upcoming games.
 
Did any of the reviews run Futuremark's API overhead test? Not against other cards, but to compare DX11 with DX12 / Mantle.

Star Swarm would also be a good one.

Something Reviewers failed to do yet again.

If i had one i'd do it, again.
 
I share your optimism for Fiji performing well under DX12. But I don't share your optimism when it comes to DX12 games. Surely there will be games out with DX12, but majority of games will ship with DX11 only. Especially games made by smaller studioes who haven't had change to get in DX12 early access programs. Though I hope I'm wrong and we see atleast 50% DX12 adaption in upcoming games.

Smaller Dev's depend on engines like Unity, There is no reason why when that and other engines get DX12 they can't port ongoing projects over.

Its not about having to start a project as a DX12 project from fresh in DX12 when its here, its simply a matter of porting what you are already working on. Edit: well, its not quite that simple.

The mechanics of it are very different from back when DX10 became DX11.
 
Last edited:

Added to list.

Why bother to make a new card when the 390X (or 290X OC'd with new drivers) is competitive with the 980 at 1920x1080, beats it handsomely at 2560x1440 and absolutely murders it at 4K?

Sorry I missed your comment.

Its more the case,a rejigged Hawaii with HBM and the Tonga tweaks,would make it much more power efficient and due to the smaller size of the HBM controller would have made the die smaller too.

This would make the chip easier to use in laptops and buy AMD some much needed marketshare in that area.

Even if Nvidia had the super high end market,AMD could get a lot of sales in the rest of the market which would help them.

However,after I made my comment I realised HBM might have more limited supply than GDDR5 which might not help.
 
pmc25, where does the 390x beat the 980 "handsomely at 1440p?

genuine question because most of the reviews I have read state this is not the case, but I know some were using the wrong drivers. Which reviews were using the correct driver?
 
Back
Top Bottom