When will we see Fury reviews?

[RussianSensation]

Damn 12k at 60fps on a single fury. Even if it graphic settings aren't on high I'm not so worried about the 64 ROPs anymore.

From your link:

"The most impressive part about the 12K AMD Eyefinity demo shown this week by AMD running the game title Dirt Rally is that they needed just one video card in the PC to push all those pixels at a playable frame rate, the AMD Radeon R9 Fury X. The frame rate was pushing 60FPS, which isn’t too bad considering what the single $649 graphics card is doing behind the scenes to push out all those pixels at an acceptable rate. AMD informed us that two Radeon R9 290X’s or a Radeon R9 295×2 get around 45-50 FPS on this exact setup."

WOW.

Also, I wouldn't worry about ROPs as it's double Tonga's and Tonga's 32 ROPs are > Hawaii's 64. TechReport once again made a giant mistake of comparing paper specs, that is despite themselves being fully aware that Tonga's pixel fill-rate is higher than Hawaii's. They should have known better from HD6970 vs. 7970 or from R9 280X vs. 285 that you cannot compare ROPs on paper in theory from different GPU architectures. Since Fury X is not based on a Hawaii architecture, TechReport's projection about Fury X's ROP bottlenecks is unfounded.

"In other respects, including peak triangle throughput for rasterization and pixel fill rates, Fiji is simply no more capable in theory than Hawaii. As a result, Fiji offers a very different mix of resources than its predecessor. There's tons more shader and computing power on tap, and the Fury X can access memory via its texturing units and HBM interfaces at much higher rates than the R9 290X.

In situations where a game's performance is limited primarily by shader effects processing, texturing, or memory bandwidth, the Fury X should easily outpace the 290X. On the other hand, if gaming performance is gated by any sort of ROP throughput—including raw pixel-pushing power, blending rates for multisampled anti-aliasing, or effects based on depth and stencil like shadowing—the Fury X has little to offer beyond the R9 290X. The same is true for geometry throughput." ~ AMD's Radeon Fury X architecture revealed

vs.

I really don't know what to say -- how do you make such MAJOR mistakes as a professional reviewer?! Wow, just wow, it's like so many of these "professional sites" can't even keep track of their own reviews. If 32 Tonga ROPs > 64 Hawaii ROPs, what happens if you have 64 Tonga ROPs vs. 64 Hawaii ROPs? According to TechReport, the same pixel fill-rate throughput. *Shakes head*.

So according to TechReport, Fury X also has little to offer beyond R9 290X when it comes to geometry throughput. Did Scott Wasson literally forget his R9 285 review or is he assuming Fury X is GCN 1.1 not 1.2? I mean come on, the data in his own reviews proves his own hypothesis wrong and if he wants to go on paper, Fury X should have close 2X the pixel fill-rate and geometry performance of a 290X (even if we account for non-linear scaling, surely 70-80% higher than a 290X).

PCPer, Legit Reviews, KitGuru, HardOCP, TechReport - so many gross mistakes and exaggerations/misconstrued data on the AMD side as of late. It's getting ridiculous.

 

[raghu78]

From your link:

"The most impressive part about the 12K AMD Eyefinity demo shown this week by AMD running the game title Dirt Rally is that they needed just one video card in the PC to push all those pixels at a playable frame rate, the AMD Radeon R9 Fury X. The frame rate was pushing 60FPS, which isn’t too bad considering what the single $649 graphics card is doing behind the scenes to push out all those pixels at an acceptable rate. AMD informed us that two Radeon R9 290X’s or a Radeon R9 295×2 get around 45-50 FPS on this exact setup."

WOW.

I would say the settings would have been handpicked to push the fill rate and put less of a burden on raw shading . We know Fiji will be a fill rate monster with that bandwidth and 64 Tonga ROPs running at 1050 Mhz. So if you chose low - medium settings which don't push shading power as much as fill rate power you could make the Fury X look better than it is actually.

Let me put it this way. At 4k the reviewers should try and strike a balance of both fill rate and shading capabilites. Ideally pick settings which let you play at 40 fps. But every reviewer will test differently. Some will chose graphics settings which push fps down to 30 (more of shading power limited) while some will choose graphics settings which are little less demanding so we will see different results. We can gain a clear understanding the various results. Thats why we should see performance across a range of reviews rather than one. :thumbsup:

 

[RussianSensation]

I would say the settings would have been handpicked to push the fill rate and put less of a burden on raw shading . We know Fiji will be a fill rate monster with that bandwidth and 64 Tonga ROPs running at 1050 Mhz.

Read the other parts of my post. I just read the TechReport's overview of Fury's architecture and there are some eye-opening statements being made in that article regarding pixel fill-rate and geometry throughout of Fury X vs. 290X. Since a lot of games are shader, texture, pixel fill-rate and memory bandwidth limited, I expect Fury to do really well in shader and texture intensive games. I think for next generation they will increase ROPs to 96-128 though. In shader intensive games such as Crysis 3, Ryse SOR, Shadow of Mordor and Far Cry 4, 980Ti is going to get slaughtered in those titles. I bet NV is working on a fully unlocked 980Ti Black Edition for late summer/early fall with the full chip and higher GPU clocks.

 

[krumme]

Thats how it is competing against nv. Its damn difficult and it calls for new meassures.
I am pretty sure the new meassure is punishment and rightly so. Thats how the game is.
So much one sided nv crap at least we will get more balanced crap. It just shows reading the reviews demand highly critical approach. More than ever.
And thanx as always rs. Its always a pleassure to read your interpretations and analysis. Better than 10 top sites. Crazy but thats how it is.

 

[Silverforce11]

Nice post RS, TR is pure fail if they don't even understand their own fill rate & geometry benchmarks. Actual ROP count isn't meaningful without know the uarch.

 

[Erenhardt]

From your link:

"The most impressive part about the 12K AMD Eyefinity demo shown this week by AMD running the game title Dirt Rally is that they needed just one video card in the PC to push all those pixels at a playable frame rate, the AMD Radeon R9 Fury X. The frame rate was pushing 60FPS, which isn’t too bad considering what the single $649 graphics card is doing behind the scenes to push out all those pixels at an acceptable rate. AMD informed us that two Radeon R9 290X’s or a Radeon R9 295×2 get around 45-50 FPS on this exact setup."

WOW.

So I was like
I bet crossfire doesn't work in this game and all they do is smokescreen comparison. So I went to the benchmarks and see this:

And now I have some tables to pick up :whiste:

 

[raghu78]

Read the other parts of my post. I just read the TechReport's overview of Fury's architecture and there are some eye-opening statements being made in that article regarding pixel fill-rate and geometry throughout of Fury X vs. 290X. Since a lot of games are shader, texture, pixel fill-rate and memory bandwidth limited, I expect Fury to do really well in shader and texture intensive games. I think for next generation they will increase ROPs to 96-128 though. In shader intensive games such as Crysis 3, Ryse SOR, Shadow of Mordor and Far Cry 4, 980Ti is going to get slaughtered in those titles. I bet NV is working on a fully unlocked 980Ti Black Edition for late summer/early fall with the full chip and higher GPU clocks.

it will be interesting to analyze the reviews of Fury X and see the trend in what is the major performance gating factor across various games. I expect Fury X to do well in the latest Cryengine games like Ryse, Evolve and texture fill rate heavy games like Shadow of Mordor and Farcry 4. R9 390X matches or beats GTX 980 in these games. Games to watch out for are GTA V, Witcher 3, Dragon Age Inquisition. Games like Project Cars are going to distort AMD Fury X's average perf across the suite as few sites like TPU and techreport include that. But people are going to be more bothered about Fury X in the average neutral games like GTA V, Witcher 3 than Project Cars.

Looking at how much R9 390X gains over R9 290X its clear to see that GCN loves memory bandwidth and Fury X has tons of bandwidth. In fact I would say excessive if we account for the color compression tech. Fury X will have >50% increase in bandwidth per sp compared to R9 290X (1.6 x 1.4 / 1.45 = 1.54). In fact even against R9 390X it will have 28% more bandwidth per shader (1.33 x 1.4 / 1.45 = 1.28). So its going to be interesting to see the performance reviews.

 

[flopper]

So its going to be interesting to see the performance reviews.

you and a whole world along

 

[desprado]

edited....

 

[Spanners]

Only 20% of DGPU is holding along but normal people know what is AMD limits and what they exactly AMD going to offer.

Are you trying to say the most recent market share numbers somehow determine who would be interested in Fiji reviews? Also you seem to be saying people who are interested aren't normal but normal people already know exactly what Fiji has to offer? You don't make any sense.

 

[Janooo]

Well 290x has an 8GB option it's just not all that common compared to the 4GB version, so really they just increased clocks, and not by any meaningful amount. You're really splitting hairs if you don't want to label that a rebadge. By that logic, factory overclocked 980Ti's with improved power delivery circuits aren't 980Ti's at all.

If we want to be precise then they are not.
They have either name prefix or suffix to them in order to differentiate them from the NV 980Ti.

 

[techguymaxc]

From your link:

"The most impressive part about the 12K AMD Eyefinity demo shown this week by AMD running the game title Dirt Rally is that they needed just one video card in the PC to push all those pixels at a playable frame rate, the AMD Radeon R9 Fury X. The frame rate was pushing 60FPS, which isn’t too bad considering what the single $649 graphics card is doing behind the scenes to push out all those pixels at an acceptable rate. AMD informed us that two Radeon R9 290X’s or a Radeon R9 295×2 get around 45-50 FPS on this exact setup."

WOW.

Also, I wouldn't worry about ROPs as it's double Tonga's and Tonga's 32 ROPs are > Hawaii's 64. TechReport once again made a giant mistake of comparing paper specs, that is despite themselves being fully aware that Tonga's pixel fill-rate is higher than Hawaii's. They should have known better from HD6970 vs. 7970 or from R9 280X vs. 285 that you cannot compare ROPs on paper in theory from different GPU architectures. Since Fury X is not based on a Hawaii architecture, TechReport's projection about Fury X's ROP bottlenecks is unfounded.

"In other respects, including peak triangle throughput for rasterization and pixel fill rates, Fiji is simply no more capable in theory than Hawaii. As a result, Fiji offers a very different mix of resources than its predecessor. There's tons more shader and computing power on tap, and the Fury X can access memory via its texturing units and HBM interfaces at much higher rates than the R9 290X.

In situations where a game's performance is limited primarily by shader effects processing, texturing, or memory bandwidth, the Fury X should easily outpace the 290X. On the other hand, if gaming performance is gated by any sort of ROP throughput—including raw pixel-pushing power, blending rates for multisampled anti-aliasing, or effects based on depth and stencil like shadowing—the Fury X has little to offer beyond the R9 290X. The same is true for geometry throughput." ~ AMD's Radeon Fury X architecture revealed

vs.

I really don't know what to say -- how do you make such MAJOR mistakes as a professional reviewer?! Wow, just wow, it's like so many of these "professional sites" can't even keep track of their own reviews. If 32 Tonga ROPs > 64 Hawaii ROPs, what happens if you have 64 Tonga ROPs vs. 64 Hawaii ROPs? According to TechReport, the same pixel fill-rate throughput. *Shakes head*.

So according to TechReport, Fury X also has little to offer beyond R9 290X when it comes to geometry throughput. Did Scott Wasson literally forget his R9 285 review or is he assuming Fury X is GCN 1.1 not 1.2? I mean come on, the data in his own reviews proves his own hypothesis wrong and if he wants to go on paper, Fury X should have close 2X the pixel fill-rate and geometry performance of a 290X (even if we account for non-linear scaling, surely 70-80% higher than a 290X).

PCPer, Legit Reviews, KitGuru, HardOCP, TechReport - so many gross mistakes and exaggerations/misconstrued data on the AMD side as of late. It's getting ridiculous.

So the contention is that Tonga ROPs are significantly more powerful than Hawaii ROPs and that reviewers aren't factoring this into their assessment of Fiji. Ok, that would seem to be the case based on the addition of Delta Color Compression, I'll grant that. However, the thing that is lacking from those 3dmark Vantage pixel fill rate tests (an almost 10 year old application, by the way) are results from Maxwell architecture GPUs. So the argument that AMD's ROPs are being under-reported by review sites is an argument that can also be made for NV's modern cards with the data *you've* presented. Now, you can cry wolf all you like about how unfairly AMD is being treated but when you use out-of-date data in order to represent one side of an argument, you seem to be cherry-picking your results to make that argument appear stronger.

Here's the proof:

The only way Fury X can produce a higher result that 980 Ti in this case is if it scales linearly with its functional unit count increase compared to its next closest relative, the R9 285. In this test 980 Ti is more than twice as fast as 290x there, but only about 20% faster than GTX 980, its little brother with 50% fewer ROPS. That tells us there are other bottlenecks here, whether they be within the GPU design, some other system bottleneck (CPU), the application being used for testing this feature, or the API utilized by the application. The bottom line is that we don't see linear scaling on this test, which is *needed* in order for this argument to hold up. Even 980 Ti's result, while impressive, is still only about 35% of its theoretical maximum rate (1GHz * 96 ROPS = 96 Gpixels/sec).

Speculate all you like, but its kind of useless without card in hand to test.

 

[railven]

So to summarize that last post:

"You're sooooooooo wrong."
"But here you're right, I'll give you that."
"But you're wrong because Nvidia."

Okay. You basically used RS's argument to use for Nvidia, where is the actual "wrong" part?

EDIT: Ah, that Ninja edit removing the whole introduction. What I get for not quoting before hand.

EDIT #2: Re-reading the edited post, yeah, I think that poster got lost some where. RS is comparing two reviews from the same site for two different products that contradict a conclusion made by the same site, ie they are not looking at their own data when making their conclusion.

This has zilch to do with Nvidia, But since you want to bring it into it, I see [EDIT: Correction, misread GTX 7xx as 9xx, my fault]

You edited out the flamebait of your post, but then made no actual counter to RS's claim.

 

[techguymaxc]

So to summarize that last post:

"You're sooooooooo wrong."
"But here you're right, I'll give you that."
"But you're wrong because Nvidia."

Okay. You basically used RS's argument to use for Nvidia, where is the actual "wrong" part?

EDIT: Ah, that Ninja edit removing the whole introduction. What I get for not quoting before hand.

RS's argument is that Tech Report's analysis of Fiji is inaccurate because they are supposedly not factoring in Fiji's use of more powerful Tonga ROPs. He then goes on to use pixel fill results from an almost 10 year old program that don't even bother to show Fiji's competition, Maxwell. By showing results across the Maxwell family of GPUs, I've demonstrated that this test doesn't have linear scaling, which would be needed in order for his contention to be true. IOW: Fiji isn't going to scale perfectly in this test because no GPU does. Every single GPU ever tested doesn't get anywhere near its peak pixel fill rate due to some unknown bottleneck.

Trying to give people the benefit of the doubt here, hence an actual explanation instead of a pointless squabble. If that's all anyone's interested in here I'll be on my way.

 

[railven]

Try actually reading this time.

Trying to give people the benefit of the doubt here, hence an actual explanation instead of a pointless squabble. If that's all anyone's interested in here I'll be on my way.

I did read, twice. You started off saying the claim was wrong (you removed that part) and then went out to say how the data was 10 years old, but it is of the 290X versus 285, which aren't even close to that old. If the app is old the same flaws apply to both test sets.

Then you used the same argument RS made to support NV with it, which is all fine, but that doesn't support your opener of "you're so wrong." I saw your original post and replied to it, sorry if you expected me to reply to your edited post.

EDIT: to your edit:
He used the data set one from site, you introduced a whole new data set. That's like trying to argue one site's power readings using another sites' methodology. Again, his claim had nothing to do with NV. He was using the same sites data points to show how that exact same site disregarded their own findings to make their conclusion on Fiji. Nothing to do with Nvidia or Anandtech or any other data/website you want to toss in there.

IE Tonga has uarch tweaks not found in Hawaii, here is your data show casing this, yet your conclusion is based on Hawaii uarch when we know Fiji isn't going to use it.

 

[techguymaxc]

I did read, twice. You started off saying the claim was wrong (you removed that part) and then went out to say how the data was 10 years old, but it is of the 290X versus 285, which aren't even close to that old. If the app is old the same flaws apply to both test sets.

Then you used the same argument RS made to support NV with it, which is all fine, but that doesn't support your opener of "you're so wrong." I saw your original post and replied to it, sorry if you expected me to reply to your edited post.

The program is 10 years old, not the data. The data RS used doesn't include Fiji's competition, Maxwell. Doesn't that seem a bit odd? Certainly not a complete representation. Not sure where the mis-communication is here. Don't assume I'm some random internet fanboy here to start a flamewar. Someone posted this on another forum I read and I saw a flaw in the argument so I figured I'd sign up here and offer a more complete data set and explain why RS' contention is relying on flawed assumptions.

 

[railven]

The program is 10 years old, not the data. The data RS used doesn't include Fiji's competition, Maxwell. Doesn't that seem a bit odd? Certainly not a complete representation. Not sure where the mis-communication is here. Don't assume I'm some random internet fanboy here to start a flamewar. Someone posted this on another forum I read and I saw a flaw in the argument so I figured I'd sign up here and offer a more complete data set and explain why RS' contention is relying on flawed assumptions.

No, it doesn't seem odd to me because he wasn't making an argument between NV and AMD. He was making an argument specifically about AMD and it's uarch tweaks between Hawaii and Tonga.

Specifically how Techspot saw results of these tweaks.

Specifically how Techspot ignored this results and made a conclusion.

Again, if the app is 10 yrs old it was used for both Hawaii and Tonga, thus the flaws apply to both.

 

[parvadomus]

The only way Fury X can produce a higher result that 980 Ti in this case is if it scales linearly with its functional unit count increase compared to its next closest relative, the R9 285. In this test 980 Ti is more than twice as fast as 290x there, but only about 20% faster than GTX 980, its little brother with 50% fewer ROPS. That tells us there are other bottlenecks here, whether they be within the GPU design, some other system bottleneck (CPU), the application being used for testing this feature, or the API utilized by the application. The bottom line is that we don't see linear scaling on this test, which is *needed* in order for this argument to hold up. Even 980 Ti's result, while impressive, is still only about 35% of its theoretical maximum rate (1GHz * 96 ROPS = 96 Gpixels/sec).

Speculate all you like, but its kind of useless without card in hand to test.

If Tonga is 20, then 2xTonga (Fiji) scores 40, and THAT is with the bandwidth limitations of Tonga we know that Fiji hast more than 2x Tonga bandwidth, plus the lower clock speed (and particularly ROPs speed) of 918 mhz vs 1050Mhz.
In the end TechReport continue to be wrong about Nv advantage in pixel fill rate...

 

[techguymaxc]

I did read, twice. You started off saying the claim was wrong (you removed that part) and then went out to say how the data was 10 years old, but it is of the 290X versus 285, which aren't even close to that old. If the app is old the same flaws apply to both test sets.

Then you used the same argument RS made to support NV with it, which is all fine, but that doesn't support your opener of "you're so wrong." I saw your original post and replied to it, sorry if you expected me to reply to your edited post.

EDIT: to your edit:
He used the data set one from site, you introduced a whole new data set. That's like trying to argue one site's power readings using another sites' methodology. Again, his claim had nothing to do with NV. He was using the same sites data points to show how that exact same site disregarded their own findings to make their conclusion on Fiji. Nothing to do with Nvidia or Anandtech or any other data/website you want to toss in there.

IE Tonga has uarch tweaks not found in Hawaii, here is your data show casing this, yet your conclusion is based on Hawaii uarch when we know Fiji isn't going to use it.

The point of using NV results is two-fold

1) to offer a more complete data set

2) to demonstrate that NO GPU ARCHITECTURE SCALES PERFECTLY ON THIS TEST

The fact that the data has more up-to-date NV cards doesn't somehow magically become invalid when discussing AMD GPUs results, they're not going to scale perfectly either so even with more powerful ROPs it's not going to smash GM200's results.

 

[railven]

The point of using NV results is two-fold

1) to offer a more complete data set

2) to demonstrate that NO GPU ARCHITECTURE SCALES PERFECTLY ON THIS TEST

The fact that the data has more up-to-date NV cards doesn't somehow magically become invalid when discussing AMD GPUs results, they're not going to scale perfectly either so even with more powerful ROPs it's not going to smash GM200's results.

1) A data set irrelevant to the statement:

2) Did you miss this:

I mean come on, the data in his own reviews proves his own hypothesis wrong and if he wants to go on paper, Fury X should have close 2X the pixel fill-rate and geometry performance of a 290X (even if we account for non-linear scaling, surely 70-80% higher than a 290X).

And again, where did he say it smashes GM200? Shoot, just re-read his post at no point did he even mention Nvidia, now did he?

 

[techguymaxc]

No, it doesn't seem odd to me because he wasn't making an argument between NV and AMD. He was making an argument specifically about AMD and it's uarch tweaks between Hawaii and Tonga.

Specifically how Techspot saw results of these tweaks.

Specifically how Techspot ignored this results and made a conclusion.

Again, if the app is 10 yrs old it was used for both Hawaii and Tonga, thus the flaws apply to both.

I'm not talking about Techspot or any other site, I'm talking about Tech Report, which RS cited by name.

I'm not making a "vs" argument, I'm using data from one vendor to illustrate that GPUs don't scale perfectly in this test, which a theoretical 2x R9 285 (i.e. Fiji) would need to do in order to beat GM200.

If Tonga is 20, then 2xTonga (Fiji) scores 40, and THAT is with the bandwidth limitations of Tonga we know that Fiji hast more than 2x Tonga bandwidth, plus the lower clock speed (and particularly ROPs speed) of 918 mhz vs 1050Mhz.
In the end TechReport continue to be wrong about Nv advantage in pixel fill rate...

The whole point is that architectures dont scale linearly in this test. That logic is flawed. GM 200 only scales 20% from GM204, despite having 50% more fill rate.

AMD uses silicon and metal to produce GPUs, just like Nvidia. Not pixie dust. The same rules apply to both companies.

 

[railven]

I'm not talking about Techspot or any other site, I'm talking about Tech Report, which RS cited by name.

I'm not making a "vs" argument, I'm using data from one vendor to illustrate that GPUs don't scale perfectly in this test, which a theoretical 2x R9 285 (i.e. Fiji) would need to do in order to beat GM200.

You are correct, it is TechReport.

But you did make it into a "vs" by first claiming he was wrong then forcing Nvidia into an argument that Nvidia had nothing to do with it.

Again, he using the data set from one specific site to show that that specific site's conclusion completely disregards the data that specific site collected.

 

[techguymaxc]

You are correct, it is TechReport.

But you did make it into a "vs" by first claiming he was wrong then forcing Nvidia into an argument that Nvidia had nothing to do with it.

Again, he using the data set from one specific site to show that that specific site's conclusion completely disregards the data that specific site collected.

You are being rather dense here. Let's go back to RS' post and analyze what he said, here is the specific claim at the end of his post:

I mean come on, the data in his own reviews proves his own hypothesis wrong and if he wants to go on paper, Fury X should have close 2X the pixel fill-rate and geometry performance of a 290X (even if we account for non-linear scaling, surely 70-80% higher than a 290X).

I provided a data set that shows 980 Ti results to give a point of comparison to Fiji's primary competition. The target is > 36GPixels/sec here. He's even admitting less than perfect scaling (albeit not nearly as conservative as they need to be, as demonstrated by the comparison of 980 Ti to 980 i.e. GM200 to GM204).

 

[Silverforce11]

The whole point is that architectures dont scale linearly in this test. That logic is flawed. GM 200 only scales 20% from GM204, despite having 50% more fill rate.

Pretty sure we know that point here. It was already stated that actual ROP count itself is meaningless without knowing the uarch underlying it.

RS's post was a correction for TR's article because they look at paper stats and see 64 ROPs, then assume its the same as Hawaii's 64 ROPs and go on to describe how it won't benefit situations in which fill-rate, AA, geometry bottlenecks. That is spinning it in a negative light, against their own data. Why do we say that? Because their own data shows Tonga's GCN 1.2 32x ROPs is superior to Hawaii's GCN 1.1 64x ROPs.

It therefore stands to reason, being a new uarch, Fury would take the improvements in GCN 1.2 at the least (or even improve upon it further). As such, Fury's 64 ROPs may actually be worth much more than 128 Hawaii ROPs.

I don't know why TR would be so clueless and miss such an obvious technical detail like that.

Btw, welcome to the AT forums.

 

[railven]

He is wrong. The argument he makes is that Fiji is going to outperform GM200 on fill rate tests. My point is that the argument is flawed because he's expecting Fiji to have perfect scaling. I demonstrated that no architecture has perfect scaling on this test and provided a list of possible reasons (hardware or software).

You must have read a different post than I did. The post you quoted, and I read, he didn't make any statement against GM. If you are inferring that, well that's your statement to argue (which you clearly are).

He even acknowledge in his very post that it isn't linear (look at Parvo's post telling you that Fiji has more than double the specs of Tonga) that it would still be at least 80%.

The Techreport conclusion was that Fiji would be bottlenecked by ROPs using the Hawaii data set. RS showed that TOnga (closer to Fiji than Hawaii) aready proves this wrong. The data was in Techreport's own articles.

At this point there is nothing to say. You are creating your own argument to fight against, if you can show me exactly where in that post RS made the following claims:

1) it will be faster than GM200

2) it scales asbolutly linearly

Then you got something to stand on. As of now, you created those arguments and are now fighting them.