AMD IQ Vs Nvidia IQ - shenanigans or something else?

[crinkleshoes]

No not you, we posted at the same time so the poster before your post so nvidia's spokesperson known as mmj.

** Comment removed **

Getting a bit bored of this now. You seem to be quite bitter about my video that was made to show a TX Vs a Fury X. Do you own either? What testing have you done? Show me your vids and I will dissect them.

I am not a professional and just a hobbyist so get off my back please.

Not meaning to interject... but that's not what I read... he's just frustrated with how AMD owners have jumped on the video as OMG AMD RULEZ VS NV type crap... when like me, I suspect he sees the difference potentially only being, in effect, a difference in the basic default software-based settings between the two companies.

I see NV ready to push the detail and open up more detail for monitor calibration... while AMD are pushing the contrast to wash out some detail in favour of more poppy blacks/colours that can be found by proper calibration.

With full calibration on the video you shared... I suspect the NV output would actually show more detail to the end user, while AMD would be lacking a little bit of detail. He may be simply terming this as an "error" on your part, without going into the details of what may have caused it.

The difference would be minimal and I'm not trying to show any favouritism... just calling what I observe myself.

mmj's expression of a similar thing, may simply have been interpreted differently from what I am trying to convey... however I see a similar message... at least in these last two posts... I haven't read further back.

I'd be happy to show you how to minimise this configuration difference... if you're interested. No worries if not.

 

[crinkleshoes]

A shot in the dark but could this be the the fault of the TEXTURE FILTERING -QUALITY option in the nvidia control panel? Either in the global or bf 4 profile? I say this because when you remove an nvidia card and reinstall it, i seem to remember that the setting defaults to the quality preset (rather than the High Quality preset).Maybe worth a look....

On the subject of the NVCP,anyone else think it seems to be getting slower and slower to respond when you first open it.Looks like your system is hanging sometimes

First thing I do for benchmarks = turn that to max performance.

As soon as I'm back to gaming... it goes straight to max quality... NV seem to like sticking that in the middle.

But by default... I don't see much of a difference there... only the contrast really in my eyes.

 

[Gregster]

Crinkle, I have been accused of gimping AF purposefully in the Fury owners thread a few times albeit said with out actually saying it, so yer I am bored of it lol

 

[crinkleshoes]

lol

 

[Bonjour]

Getting a bit bored of this now. You seem to be quite bitter about my video that was made to show a TX Vs a Fury X. Do you own either? What testing have you done? Show me your vids and I will dissect them.

I am not a professional and just a hobbyist so get off my back please.

Keep it up, dude.

Many people are appreciative of your efforts. It's very time consuming what you're doing, particularly swapping between the two cards. Don't let one or two bad eggs get you down.

 

[Gregster]

Keep it up, dude.

Many people are appreciative of your efforts. It's very time consuming what you're doing, particularly swapping between the two cards. Don't let one or two bad eggs get you down.

Thanks and if I am guilty of anything, it is being a noob with a capture card I will get better though.

 

[Mauller]

Thanks and if I am guilty of anything, it is being a noob with a capture card I will get better though.

Which card did you test first greg? the FX or the TX? just wondering if it could be an issue with BF4 settings in the end. If it detects the GPU and alters anything internally etc. Could be an issue where you wanted to clear the settings folder before using the second card.

Think above would apply if you didn't alter any settings between card changes. Assuming they should remain the same etc when you swap cards.

 

[Freddie1980]

Wow Gregster has changed since his suspension.

 

[arc@css]

I wanted to point this out many times since first seeing this thread, but held back because of afraid of being accused of trolling. Glad that someone not me pointed this out...

After all, graphic cards reviews these days is all about FPS vs FPS, and only on the odd occasions would things like framepacing/smoothness/stuttering be looked at, nevermind image/graphic quality...

+1

I remember the days of old when you would see a definite IQ difference when upgrading (even the same brand). The good ole pipe line days.

 

[shankly1985]

Waste of time mate, I've seen about 10 such posts across various forums now and if it were a real issue the top media outlets would have been all over it by now.

AMD users will cling to Gregster's video because it's all they have to promote Fury over 980Ti regardless of it being an obvious issue (intentional or not) at Gregster's end, he's had some fantastic publicity thanks to the desperation of shankly etc spamming it everywhere and he'd look stupid to admit to a testing fault now, especially considering the basic ones he's already had to admit and the obvious black crush on Fury's side that I've been pointing out and that AMD users don't want to talk about (which is probably another fault at Gregster's end as well).

Bah, Bah, bah..

I have made this clear right from the get go that this could very well be a simple driver setting or a driver bug on battlefield 4..

I have also pointed out a very good reason why amd blacks might seem abit more crushed.. I'll tell you again. Greg recorded with pvr by hdmi, amd by default hdmi is set to ycbcr444, if the pvr is recording at rgb full or even rgb limited then it will make the brightness levels way off..

I only posted it elsewhere to get more people's advice on the matter because Greg said he didn't want to get involved. I accepted that and went elsewhere to find the answer.

So let's get fact straight. Ignore colour, gamma brightness or even contrast because none of these matter... Zero none

What matters is the draw distance details is missing completely, I'll say it again missing completely.... Nothing to do with calibration of colour etc.

Mah

 

[Kaapstad]

Getting a bit bored of this now. You seem to be quite bitter about my video that was made to show a TX Vs a Fury X. Do you own either? What testing have you done? Show me your vids and I will dissect them.

I am not a professional and just a hobbyist so get off my back please.

From what I have seen you seem to be doing very well with the videos, if you have got any more they would be welcome.

 

[TNA]

From what I have seen you seem to be doing very well with the videos, if you have got any more they would be welcome.

Agreed. I enjoy watching his videos (the voice), to me it comes over as being genuine, even if I may at times not agree with something being said, I know he is likely telling the truth from his perspective.

It is nice to watch reviews from people like gregster, at least it is less likely that he is in contact with one company or the other, so we can count on his opinion rather than someone worried to **** them off and miss out on a free review sample in the future.

Keep up the good work Gregster fella

As for mmj, lambchop and dp, give it a rest

 

[Gregster]

Agreed. I enjoy watching his videos (the voice), to me it comes over as being genuine, even if I may at times not agree with something being said, I know he is likely telling the truth from his perspective.

It is nice to watch reviews from people like gregster, at least it is less likely that he is in contact with one company or the other, so we can count on his opinion rather than someone worried to **** them off and miss out on a free review sample in the future.

Keep up the good work Gregster fella

As for mmj, lambchop and dp, give it a rest

Thanks and I am always appreciative of positive comments like this. The last bit about mmj, chop and d.p is harsh though and I am interested in what people think and say but not keen on being told I deliberately missed things out. This is the thread to discuss it.

 

[D.P.]

+1.

It should, it has 45% more Shaders and about 70% more memory bandwidth yet at the moment it seems to be struggling to get 30% past a similar clocked Hawaii.

There is a lot of performance missing from the architecture.

I think AMD are putting a lot of effort into Win 10 Drivers, Win 8.1 has taken a bit of a redundant position and thats fine, for Gamers Win 8.1 is about to become redunadnt.

In time and probably not much time i think it may grow very well into Win 10.

45% more shaders but much of the rest of the design is the same, e.g. Rops, TMUS, Tesselation engines. The extra streams are also still only in 4 shader engines. Now there have been additional tweaks here and there some as color compression and refinement of the geometry engines but nothing big so I certainly wouldn't expect 45% more performance in real world gaming. 30% clock for clock on average across different games seems reasonable to me, with bandwidth or pixel shader limited games getting nearer 40%. Then there is the issue that more compute units will require better load balancing in order to scale, diminishing returns are all too common.

Remember removing one bottle neck simply let's you hit the next bottle neck.
The GCN architecture has really been pushed to its limits, it can't handle more than 4 shader engines, can't have more than 1 geometry engine per shader engine and the rop count is fixed to shader engines etc. And that is why the best AMD could do was add co outer units to the fixed number of shader engines and use their Tahiti upgrades. It probably wasn't possible to move to a new GCN architectuse that supports more shader engines for example while stuck on 28nm. A future iteration with 6 shader engines would increase geometry rates 50% and would liekly make slacking efficiency of the compute units easier, a design change to have 2 geometry engines per shader engine would probably be effective in keeping up with nvidia.


Anyway, looking at theoretical number is mostly worthless,s Fury does hit its theoretical performance in the synthetic benchmarks which strongly indicates there aren't serious driver issues. Other bottlenecks and diminishing returns likely explains the performance characteristics.

The TX doesn't perform an exact percentage faster than the 980 from the theoretical difference, does that mean there will also be big gains in Titan and 980Ti performance with drivers? Or do we simply accepts things don't scale linearly.

 

[humbug]

45% more shaders but much of the rest of the design is the same, e.g. Rops, TMUS, Tesselation engines. The extra streams are also still only in 4 shader engines. Now there have been additional tweaks here and there some as color compression and refinement of the geometry engines but nothing big so I certainly wouldn't expect 45% more performance in real world gaming. 30% clock for clock on average across different games seems reasonable to me, with bandwidth or pixel shader limited games getting nearer 40%. Then there is the issue that more compute units will require better load balancing in order to scale, diminishing returns are all too common.

Remember removing one bottle neck simply let's you hit the next bottle neck.
The GCN architecture has really been pushed to its limits, it can't handle more than 4 shader engines, can't have more than 1 geometry engine per shader engine and the rop count is fixed to shader engines etc. And that is why the best AMD could do was add co outer units to the fixed number of shader engines and use their Tahiti upgrades. It probably wasn't possible to move to a new GCN architectuse that supports more shader engines for example while stuck on 28nm. A future iteration with 6 shader engines would increase geometry rates 50% and would liekly make slacking efficiency of the compute units easier, a design change to have 2 geometry engines per shader engine would probably be effective in keeping up with nvidia.


Anyway, looking at theoretical number is mostly worthless,s Fury does hit its theoretical performance in the synthetic benchmarks which strongly indicates there aren't serious driver issues. Other bottlenecks and diminishing returns likely explains the performance characteristics.

The TX doesn't perform an exact percentage faster than the 980 from the theoretical difference, does that mean there will also be big gains in Titan and 980Ti performance with drivers? Or do we simply accepts things don't scale linearly.

Actually i think the tessellation engine is different.

Like Fiji is to Hawaii Maxwell is no different to Kepler, Maxwell gains its performance from Texture and Delta Colour Compression, Fiji has the same technology.

The same argument your using here applies to Maxwell vs Kepler and yet the 3072 Shader GM204 looks like its about 70% faster than the 2880 Shader GK110, or at least it does in games that are usually benchmarked.
It doesn't gain much at all in synthetic benchmarking, only to the Shader and clock rate difference.

Now, the thing is they can't tweak drivers to reduce IQ and gain performance that way in FutureMark, it would void the result.

Add that to Gregsters obvious IQ difference between the Fury-X and Titan-X, Maxwell gaming performance starts to look like (Driver optimisation)

 

[mmj_uk]

Getting a bit bored of this now. You seem to be quite bitter about my video that was made to show a TX Vs a Fury X. Do you own either? What testing have you done? Show me your vids and I will dissect them.

I am not a professional and just a hobbyist so get off my back please.

Your video is the only source in existence at this point which shows such an anomaly so I'm not sure what there is for me to prove? there have already been at least a dozen posts throughout sites today showing your video to be unique in its rendering of BF4 on multiple cards throughout the NVidia range, no review site has rushed up an article to gain a massive exclusive.

Hobbyist or not if you're publically promoting a Youtube channel called "Gregster reviews" and posting videos with obvious flaws that give false impressions to buyers on a shops' forum you are going to come into some criticism. If I were going to promote a "review channel" I would fully expect criticism myself, if you don't want it then don't post links to them on public forums.

The state of play so far:

- Videos were initially recorded at completely different bitrates making Fury appear better quality, coincidently(?) said video was initially posted in a Fury thread.
- Titan has missing AF or LOD issues in BF4 that nobody else on multiple forums can seem to replicate.
- Fury has black crush but those AMD supporters who are pushing the Titan anomaly are just ignoring that because it doesn't suit.
- Gregster has far more views on his channel due to all of the above.

 

[mmj_uk]

Actually i think the tessellation engine is different.

It's the same GCN1.2 setup as Tonga, probably helped somewhat by the additional grunt and coupled with more aggressive 'optimisations' ie. AMD are being more proactive overriding the tessellation complexity to lower factors via drivers to relieve bottlenecks in their geometry/tessellation engine.

 

[shankly1985]

Tbh Gregster knew nothing about the draw distance, it wasn't till boom pointed it out.

The video was just showing performance.

 

[humbug]

It's the same GCN1.2 setup as Tonga, probably helped somewhat by the additional grunt and coupled with more aggressive 'optimisations' ie. AMD are being more proactive dumbing down the complexity to lower factors via drivers to relieve bottlenecks in their geometry/tessellation engine.

Yes and Tonga despite 33% Less Bus 8% lower clock rates and 15% less Shaders is closer to the 7970/280X performance than it is to the similar but still bigger 7950/280, sometimes beating the 280X convincingly.

 

[Kaapstad]

Thanks and I am always appreciative of positive comments like this. The last bit about mmj, chop and d.p is harsh though and I am interested in what people think and say but not keen on being told I deliberately missed things out. This is the thread to discuss it.

I don't think you have missed anything out, I also think you have been very professional in your review.

More of the same please.