Middle Generation Console Upgrade Discussion [Scorpio, 4Pro]

[Kaotik]

Really?! I was always under the assumption it was a X1800 XL. What was it based upon if you don't mind sharing? I read through numerous articles back then, T-Report, Ars, Anand, Semi, [H]ardocp etc. Xenos was always described as unique but I always thought it was based on something in the ATi 1000 series.

It's the offspring of the fabled R400 project, which evolved into R500 into C1 aka Xenos (or was it Xenon? can't remember).
It was ATIs first unified shader GPU that got into production, it was something "midway" R300 and R600.

Someone once said it's performance was similar to X1800(XL?), which might have misled you

 

[Kaotik]

Okay, I wasn't aware of that.
But does that mean anything? Did it mean anything in the past?

At least ISA changes in the past

 

[MrFox]

I know nothing of GCN versions, but is there any chance of having more operations per clock with an ISA change, like wider instructions? Has this ever happened in the past?

 

[eastmen]

It's a given, 64 for cut(56-64 CU), 96 for full(88-96 CU) Vega.

How, exactly? That wouldn't even be 6TF gpu anymore
MS is hardly going to pay for custom(not even semi-custom like X1's) chip.

Because the system doesn't have to be just 6Tflops. It just has to be at least 6Tflops.

MS isn't going to say oh wait this is 6.4 tflops fuck it , we can't use it we can only do 6tflops. MS set a floor of what they have to at least hit. I am sure their goal is for more . Also the rops and tmus would affect flop count but would make a big difference in performance at higher resolutions.

 

[Proelite]

The earliest Xb1 target specs from MS called for a 1.2 tf GPU. It ended up being 1.31 tf in the end.

I expect Scorpio to be in between 6. 1 to 6.6 simply because probablity states that a properly balanced GPU hitting a whole number is pretty low.

By no means am I insinuating that there is a green and orange pill.

 

[EvenBiggerAl]

Have MS confirmed the GPU alone will be > 6Tflops? I've seen a quote from Spencer saying 4.5x the XB1 (i.e. 5.9 Tflops which works out at almost exactly 56CU @ 825mhz).

If it were greater than 6Tflops I would have thought they would state that explicitly (even if it was just 6.1Tflops) whereas if it was 5.9Tflops then they would likely round that up and state 6Tflops

 

[MrFox]

The earliest Xb1 target specs from MS called for a 1.2 tf GPU. It ended up being 1.31 tf in the end.

I expect Scorpio to be in between 6. 1 to 6.6 simply because probablity states that a properly balanced GPU hitting a whole number is pretty low.

By no means am I insinuating that there is a green and orange pill.

I would consider the marketing suits theory: AMD presented MS the same config they did with Sony from the rumors, 384bits GDDR5, pulling off 5.5TF with a conservative clocking. They could have decided it was worth sacrificing a bit of yield and power consumption to reach 6TF for marketing reasons.

 

[BRiT]

Have MS confirmed the GPU alone will be > 6Tflops?

CPUs do not add significantly to Flops.

 

[kalelovil]

No, it isn't. AMD's OpenCL drivers already confirmed that Vega is gfx ip 9.x (which was claimed ages ago at linkedin profile of ex(?)-AMD employee), while Polaris is gfx ip 8.x just like tonga and fiji

Now that it is out, has anyone confirmed that is true of Polaris?

 

[EvenBiggerAl]

CPUs do not add significantly to Flops.

I know. I was just considering it from a marketing POV for MS. 6TF is a good number to clear so if the GPU were to be 5.9TF then they can either round that up and/or include the 120-150Gflops of the CPU and market it as > 6TF

 

[scently]

I know. I was just considering it from a marketing POV for MS. 6TF is a good number to clear so if the GPU were to be 5.9TF then they can either round that up and/or include the 120-150Gflops of the CPU and market it as > 6TF

Because they stated it in the unveil video that its the gpu.

 

[Kaotik]

Now that it is out, has anyone confirmed that is true of Polaris?

The same OpenCL driver confirms Polaris is 8.x (though it does put it in different "group" than Tonga & Fiji, probably due updated hardware side even though ISA is the same)

 

[Proelite]

I don't think these minute fraction of a teraflop is going to matter much in the system's perf. Only useful for when fanboys need to lay them out on the table and measure in order to win some envy contest.

Spoiler

Which I enjoy seeing.

 

[upnorthsox]

Scorpio prediction
(48 CUs @ 950 mhz = 5.836 tflops) + (8 @ 2.1 ghz = .168 tflops) = 6.004 tflops

 

[MrFox]

I don't think these minute fraction of a teraflop is going to matter much in the system's perf. Only useful for when fanboys need to lay them out on the table and measure in order to win some envy contest.

Which I enjoy seeing.

True.

I expect Scorpio to be in between 6. 1 to 6.6 simply because probablity states that a properly balanced GPU hitting a whole number is pretty low.

What I enjoy seeing is imaginary fractions of teraflops... Those are the best kind of teraflops.

The media will use the numbers, imaginary or not, in a way that will influence sales, 5.9 is much lower than 6. It has no influence on performance, but sales take a hit once the media use it.

 

[GrimThorne]

Well some of this is my own speculation on the matter:
The R500 IIRC was a project that was initially meant for replacing the R400 in the PC space (which by itself was a rather small evolution over R300, meaning ATi's resources had been centered elsewhere), but was put aside circa 2004. Supposedly, back then the transition to unified shaders and an effective use of eDRAM in Windows could be a huge hassle driver-wise, so instead they sold the VEC4+Scalar architecture to Microsoft, who was looking for something with a lot of potential inside a closed environment. Then R520 and R580 came with non-unified shaders and by late 2006 they figured Terascale (VLIW5) would be more appropriate for the desktop.
Xenos' VEC4+Scalar was still used for ATi's Imageon line of GPUs for handhelds (back then almost exclusively used by Intel for their ARM-based SoCs called XScale), specifically for their next-generation OpenGL ES 2.0 Imageon called Z430, or Mini-Xenos. Qualcomm used ATi's Imageon for their MSM7xxx SoCs and the Z430 went into MSM7227 and the first-ever mobile SoC with a 1GHz CPU called Snapdragon.
When Qualcomm bought the handheld GPU division from ATi (by then already a part of AMD IIRC), they changed the Z430 name to "Adreno", which is actually an anagram of "Radeon".

Adreno. I would have never figured that one out for myself so there goes my Double Jeopardy points for the day. Thanks for taking the time out to both explain and detail this for me. So much went into Xenos, I enjoyed the deep dives but I didn't know the backround story behind it. Thanks again, most appreciated.

It's the offspring of the fabled R400 project, which evolved into R500 into C1 aka Xenos (or was it Xenon? can't remember).
It was ATIs first unified shader GPU that got into production, it was something "midway" R300 and R600.

Someone once said it's performance was similar to X1800(XL?), which might have misled you

There were also claims that it was based on the X1900 series as well. But I do remember name C1/Xenos, I just can't recall where I read it. It's been awhile since 2005.

 

[Kaotik]

There were also claims that it was based on the X1900 series as well. But I do remember name C1/Xenos, I just can't recall where I read it. It's been awhile since 2005.

One can claim anything, reality is it isn't "based" on any consumer GPU or even architecture (even if R400 was meant as such originally)

 

[kalelovil]

Scorpio prediction
(48 CUs @ 950 mhz = 5.836 tflops) + (8 @ 2.1 ghz = .168 tflops) = 6.004 tflops

Presuming it uses Zen cores, that CPU tflop figure should be doubled.

 

[goonergaz]

Because they stated it in the unveil video that its the gpu.

They stated SoC not GPU @55 seconds then right at the end there's mention of 6TF GPU - so mixed messages?

Or one is stating fact the other just his interpretation )who is who because @55 seconds he's saying 'we gave the SoC 6TF' whereas at the end the guy says 'this thing has a 6TF GPU' - my interpretation would be @55 secs is factual and the guy at the end is the sound-bite they want you to remember as he reels off the top-line specs.

 

[Rangers]

If it uses zen cores then yikes! They will have outdone themselves.

Hadn't thought about it much, just assumed Jaguar cores. I'd still lean that way.