Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Jun 24, 2015.
It would need to be ported over to DX12. Games may get some improvements from WDDM 2.0 though.
If AMD made the best CPU's they would be used but they don't. My 8350 can do a few things well but even Intel CPU's that are a few years older are better at gaming and don't bottleneck like the 8350 does.
I'm not big on O.C.ing, but I'm at a stable 4.3 ghz using the Asus auto clocking feature .
I'm not saying that a pure AMD system will have different results, but it would help us Tin foil hat folks see how the card will scale in a AMD system. .
Anyone who thinks DX12 will magically favor AMD and give them the performance crown is surely deluded.
I have had AMD cards in my system for the last 10-15 years.
Have decided to go Green this time and just ordered the MSI GTX 980Ti Gaming from CCL.
PhysX and the cooler more efficient GPU helped bring me over.
Poor drivers from AMD were also a "MAJOR" part of my decision to change over as well as the Fury not quite being good enough.
If DX11 has anything to say about it, not great.
It will benefit, but in DX11 and especially on AMD side, multithreading does not matter much. Therefore intel gives benefit from high "single thread performance" (IPC +good clock).
But when you take DX12, total extractable compute power of CPU starts to matter. And there i5 @3.8GHz is weaker than FX-8350 @4.3GHz.
Poorly mate, poorly. At least till AMD rewrites that single threaded part of driver. (probably decade old code)
I must admit I am not knowledgable on the subject, but Mantle was patched to some games.
What would be so hard to add a patch for a DX12 path, even if only partial ?
DX12 scales well up to 4 cores on Star Swarm and DX12.
At the moment I don't see AMD benefiting any more than Intel.
Wow whatta review. 38 pages! Do we have the record?
Damn shame about idle noise. Ruining this cards major advantage over 980 Ti.
Actually frametimes are not that good.
They are worse than 980.
But honestly, this is kinda expected. Many games depend on game specific optimizations, and Fury X being brand new is missing those.
Fiji will need a lot of love from driver team in upcoming months to compete on its own against GM200.
The only reason why these prices (same as 980 Ti) might stick for a while, are supply issues.
I just can't find it in me to suggest buying the Fury X over the 980Ti. In that sense what's the point? Unless unlocked voltage control turns it into an overclocking monster I just don't see one.
It is actually up to 6. You'll not see another gain in case your GPU becomes bottleneck.
And remember that this is very simple demo. If you utilize 4~6 threads by DX12 library, what kind of compute power you expect to be left for main game thread and its spawns?
It gains 0.01 FPS from 2 cores?
sure it does, in synthetic and irrelevant benchmarks. :nerd:
From the reveal, I was expecting more overclocking headroom.
Even a slight voltage increase will put out serious heat.
Look at the 390X power consumption when overclocked.
I mean even 1200 would be probably the limit, if not too much, for the cooler.
edit : LN² might help. Wonder who will be the first ?
It runs 49C stock voltage. I don't think cooling will be the problem.
Shame it's locked to reference.
From Hardware.fr, their FLIR photo of the back of the card, without the backplate.
Do you really expect me to spoon feed you reality of GPU bottleneck in between 2 and 3 cores in that particular test and settings?
If you want to see how CPU scales when GPU is not bottleneck, you run at low resolution, low details. They did run it on Extreme. If you already push into GPU more workload than it can handle you'll not see how it improves with pushing more.
That is why 3DMark API overhead test uses lower resolution and pretty primitive object pushed to GPU again and again.
How about the rest of the charts?
Looks like higher ROP count does matter, in one review AMD said 64 is enough with new compression and HMB,.
Guess not, but its still a little impressive vs NV that needs 96 ROPS to take the lead.
+ fast at 4K resolution, here its TitanX competitor, if 30-60fps means fast :d
-/+ performance could be better, although it still rivals 980Ti and TitanX 2x more $$.
- average OC with review samples.. I expected at least 1200mhz+:nerd:, not max 1160-1175mhz or 1150mhz or 1140mhz, 1130mhz,... Although AMD said it may vary, hopefully all these are lower bind samples.
Price, well both gpus Amd & Nvidia should have been in 600€ area max, not 700-750€..