NGGP: NextGen Garbage Pile (aka: No one reads the topics or stays on topic) *spawn*

[Shifty Geezer]

If Orbis had 23 and a half CU, I guess we could say it has almost 24CU.
I don't see 50% as being almost 100%, maybe it's just me.

What about 1x is almost 2x, If it's equal it's almost twice.

Mathematically when rounding, 1.5 => 2. This is what kids are taught in school. Personally, I teach you should round to the nearest half for accuracy, not the nearest whole. 1.1 = 1.0, 1.4 = 1.5, 1.6 = 1.5, 1.8=2.0. Don't be surprised if people do use large roundings though - that's how they've been taught to think.

 

[Blazkowicz]

I only 0.25x agree with you.

Besides, Orbis is 150% the performance of Durango, so it's almost 200% faster. Thus, Orbis is 3x as fast as Durango

 

[upnorthsox]

I only 0.25x agree with you.

Besides, Orbis is 150% the performance of Durango, so it's almost 200% faster. Thus, Orbis is 3x as fast as Durango

Actually 3.5x which is almost 4x so Orbis could be 4x as fast as Durango.

 

[Gubbi]

Anyone who had physics in high school would know to round numbers to three significant digits (doesn't work with year/dates in history though)

Cheers

 

[Shifty Geezer]

Anyone who had physics in high school would know to round numbers to three significant digits (doesn't work with year/dates in history though)

Humanbeings work to different degrees on accuracy depending on context. Ordinarily, outside of science, 2 sf is enough (hence the invention of percentages). In estimating and approximating and talking general terms, 1sf is often used. Which personally I don't think is good enough and recommend the half values too. There's no real point in differentiating between 11% and 12% faster when talking about general power differences, but the difference between 5% and 10% is definitely worth observing.

Anyway, Lottes didn't. He made a broad generalisation and continued to discuss that generalisation as if it were a data point, which is hardly a great basis for a technical comparison of two (unknown) consoles!

 

[Blazkowicz]

Anyone who had physics in high school would know to round numbers to three significant digits (doesn't work with year/dates in history though)

Cheers

Not if you have a measure or data with less than three significant digits.

 

[mrcorbo]

Anyway, Lottes didn't. He made a broad generalisation and continued to discuss that generalisation as if it were a data point, which is hardly a great basis for a technical comparison of two (unknown) consoles!

The fact that the blog post was based on old information explains quite a bit. It looked very odd initially how he listed, in great detail, all of the advantages GCN brought to Orbis and then for Durango...nothing.

And the performance rounding was also jarring. I'd have expected a graphics programmer to know how much a lack of precision can distort a view.

 

[Shifty Geezer]

Well he has no vested interested. Just some vague ramblings, the same as the rest of us only somewhat less well informed because he used SemiAccurate at a reference.

 

[MrFox]

It's interesting how we (including gamers, journalists, coders) are switching between "number of times faster" and percentage of additional power, depending on how close they are. Anyone comparing with previous gen or even the wiiu will say how many times faster it is, and it's only when it's below 2x that we switch to percentage. Does it mean we intuitively care only about 1 sf? (that's possibly my brain fart of the day, thank you)

 

[Butta]

Mathematically when rounding, 1.5 => 2. This is what kids are taught in school. Personally, I teach you should round to the nearest half for accuracy, not the nearest whole. 1.1 = 1.0, 1.4 = 1.5, 1.6 = 1.5, 1.8=2.0. Don't be surprised if people do use large roundings though - that's how they've been taught to think.

So that means Durango 1.2 = 1.0 and Orbis 1.8 = 2.0
So Orbis is double Durango

 

[Shifty Geezer]

So that means Durango 1.2 = 1.0 and Orbis 1.8 = 2.0
So Orbis is double Durango

That's the broad comparison some have sadly been making.

 

[Sisco]

Something is really fishy about the next xbox gpu. If the rumours are true, it will be a very expensive part of the console.

 

[french toast]

Something is really fishy about the next xbox gpu. If the rumours are true, it will be a very expensive part of the console.

How so?..much less than last generation is it not?

 

[kots]

Why should be expensive ? It's a GPU of middling performance .. are you talking about the "extra parts" ? I don't believe they exist or they are at any importance....
If those "extra parts" are expensive , why don't they just get a better GPU?
Nobody can't answer this question.

 

[Sisco]

That's what i want to know. what's the point of those extra parts, when they could go with a better gpu??? It's weird.

 

[AlphaWolf]

How big are they?

A lot of engineering a console is to make efficient use of thermal limits and build costs. If MS thought they'd get similar or better performance by allocating the die space of the embedded memory and move engines to more CU's, they'd probably have done that. MS apparently wanted 8GB of ram which makes gddr5 not feasible, so part of the design is to make up for the bandwidth deficit.

You can bet MS explored other builds and have chosen one because it offered the best performance for their wants and needs. You can also bet that they looked at some data on gpu usage.

 

[Acert93]

You can bet MS explored other builds and have chosen one because it offered the best performance for their wants and needs. You can also bet that they looked at some data on gpu usage.

The bold is the $1M dollar question. As BK has indicated the old model of core gamer first and casual gaming as a natural consequence may not be the order of priority. Wii and Kinect may ahve something to do with that.

 

[Scott_Arm]

The bold is the $1M dollar question. As BK has indicated the old model of core gamer first and casual gaming as a natural consequence may not be the order of priority. Wii and Kinect may ahve something to do with that.

I believe what he said was correct. Best performance for their needs and wants. As in, best performance for a box with 8 GB of RAM and under $x. If the DMEs could have been replaced with better CPU, GPU in their price/performance range, they would have done that.

 

[Sisco]

If there's a big diference in games like COD, FIFA, GTA, Microsoft will be in some serious deepshit. How will they justify live gold?

 

[Proelite]

I believe what he said was correct. Best performance for their needs and wants. As in, best performance for a box with 8 GB of RAM and under $x. If the DMEs could have been replaced with better CPU, GPU in their price/performance range, they would have done that.

Could it be possible that there is more and more diminishing returns on the GPU taking in consideration the ~70GB/s bandwidth and 5GB of DDR3?

I believe they built a GPU solution that made maximal use around >=5GB of DDR3.