Wrong information from AnandTech for GPU

[Plamensito]

Hello

I want to say that the information you give AnanTech for Apple GPU and Tegra GPU is not at all true.



AnandTech assert that A6X has 71.6 GFLOPS???

Your information is wrong, because it would appear that Mali T604 is more powerful than the PowerVR SGX554 MP4???

Mali T604 - 17 x 4 x 0.533MHz x 2 = 72 GFLOPS

Apple GPU SGXMP that there is one scalar:

4x2 + 1 = 9 GFLOPS

A6X :

Now 32 x 4 x 0.280MHz x 9 = 80 GFLOPS


By your ridiculous logic goes that the T604 is more powerful than A6X -hahahahs

AnandTech Says:

Nvidia Tegra 4 Wayne is a 74 GFLOPS. This is wrong!

Tegra 4 Wayne:

48 vertex units + 24 pixel shaders x 0.672MHz x 2 = 96 GFLOPS

You are amateurs!

Regards, Plamensito - 359gsm.com (SoC & GPU) & Wikipedia

http://www.359gsm.com/forum/viewtopic.php?f=127&t=13396

http://www.359gsm.com/forum/viewtopic.php?f=127&t=13134&start=10

 

[3DVagabond]

Is this directed at a person or a particular article? Just leaving it here like this with no further reference, it's going to die a quick death.

 

[Grooveriding]

http://www.youtube.com/watch?v=l1dnqKGuezo

 

[Torn Mind]

The article writer might not be a forum participant.

 

[notty22]

http://www.youtube.com/watch?v=l1dnqKGuezo

A video response. http://www.youtube.com/watch?v=HKmZhfRmtng

 

[BallaTheFeared]

You are amateurs!

No frame time graphs or video output capture :thumbsup:

Fishy fishayyy!

 

[WaTaGuMp]

I feel smarter already.

 

[DominionSeraph]

AMD and Nvidia both top 3 TFLOPS. Red and Green both beat Team Fruit.

 

[ViRGE]

1) I'm not going to call you wrong, but no one counts the scalar unit on PVR Series 5XT. It's right up there with G80's missing MUL. It can only be used in a handful of scenarios, and while important it's not usable in the same way that the vector units are. Note that AMD's scalar unit on GCN hardware is also not counted when computing theoretical performance.

2) If you're going to complain, at least get the number of units right on Tegra 4. You've swapped the number of vertex and pixel shaders. This is important because the pixel shaders don't have nearly the precision of the vertex shaders (FP24 versus FP32).:|

That said, you appear to be right about the quoted speed for T4 being wrong. It looks like AT copied that chart from their first T4 article, which was before T4's clockspeeds were known and they used an assumed 520MHz (the same as T3).

http://www.anandtech.com/show/6666/the-tegra-4-gpu-nvidia-claims-better-performance-than-ipad-4

http://www.anandtech.com/show/6787/...ture-deep-dive-plus-tegra-4i-phoenix-hands-on

Ultimately you need to email Anand since he would have written the article. He's unlikely to see your post here.

 

[Face2Face]

You are amateurs!

Calling Anand an amateur? Is this guy For Serious?