Scarlett SOC leak spotted? Code name Flute? And how does it compare to PS5 Gonzalo APU?

AMD Flute SoC Rumored for Xbox Project Scarlett Sees First Leak

An AMD system-on-a-chip (SoC), codenamed Flute, has popped up in the UserBenchmark database (it's since been removed). This mysterious SoC could power Microsoft's next-generation gaming console, codenamed Project Scarlett, which Microsoft has said will use AMD Zen 2 architecture and Navi graphics.

AMD is pretty busy nowadays producing custom-designed chips for Microsoft and Sony. The chipmaker is reportedly producing the AMD Gonzalo SoC for the PlayStation 5, and with AMD Flute now making an appearance, the new rumor is that the SoC will power Project Scarlett.

The leaked AMD Flute sample is presently identified with the "100-000000004-15_32/12/18_13F9" codename. It's a bit bizarre that the OPN (Ordering Part Number) is included in the codename. On the other hand, it could mean that it's a qualification sample (QS).

AMD SoC Rumored Specs*

AMD Flute is an eight-core, 16-thread processor, that seemingly comes with 1.6 GHz base clock speed and a 3.2 GHz boost clock. There is no official information on whether AMD Flute is based on the Zen, Zen+ or Zen 2 microarchitectures. Microsoft has previously stated that Project Scarlett uses a Zen 2 processor.

The AMD Flute sample features the 13F9 PCIe ID, which alludes to the Navi 10 Lite silicon. As we've recently seen in a Linux display driver, there are a couple of Navi 10 Lite variants in the wild. The AMD Gonzalo qualification sample has the 13F8 PCIe ID and purportedly runs with a 1.8 GHz iGPU (integrated graphics processing unit) clock. Unfortunately, the UserBenchmark entry didn't list the iGPU speed for the AMD Flute part before it was removed.

https://www.tomshardware.com/news/amd-flute-soc-xbox-scarlet-benchmark-leak,40000.html

What do you make of it?

If it turns out to be accurate are you satisfied or disappointed?

No clue, but i think 1.6ghz seems low only cause it's a 1.

lol makes sense.

I'm only here for Ron and his million pie charts and graphs.

I don’t care. I just like playing video games.

1.6 base clock is wild. They’ve learned nothing...

@cainetao11: The SNES has one of the best libraries till this day.

Same. The specs can be talk about once these next-gen consoles are release.

While I wouldn't know what to make of this I imagine people who are designing these systems would have more insight to its meaning. And if you are MS trying to play your cards close to your chest and this info got leaked then maybe their competition knows how maybe to tweak plans on their console to gain a performance edge. Then again, if I wanted to throw off my competition I might want to stage a leak to have them assume they only need to be marginally better then surprise them later when it's too late, but I'd probably be giving them too much credit to assume they'd go to such lengths.

Honestly don’t really know what to think of that. I guess I’ll just wait and see what the games will look and play like. If I like what I see then I’ll be satisfied.

@goldenelementxl: it would run at 3.2 for games.

@r-gamer: No, 3.2ghz is the boost clock. In AMD speak that means single threaded workloads. 0% chance you'll ever see all cores 3.2ghz. Games will run 2.2ghz - 2.5ghz max.

There is nothing linking this information to any console outside of just speculation. Besides, it next gen is going to be the smallest generational jump outside of WiiU to Switch.

Right. The single core boost is 3.2, which isn't great anyway. Hell, heavy threaded games could run much close to the 1.6 base clock. With the frequencies we're seeing with Zen2, I am SHOCKED that these are the clocks we're getting.

Flute is likely Gonzalo.

Gonzalo is the APU, Flute is the platform as a whole

And the graphics score on Fire Strike is rivaling a RTX 2080 supposedly

Really? You expected more despite the power constraints?

@Pedro: The 3700X is a 65W tdp chip with a 3.6 base and a 4.4 max boost. I thought we were getting a cut down version of that in the next consoles, but not this cut down. The Xbox One X tdp is around 245w. The jag CPU is clocked at 2.3 and is far less efficient than Zen 2. Why not keep the base closer to 2.3 next gen? 1.6 just seems nuts to me.

Who cares Giovela/Carlos you're going to play most games on your high end PC again like you did over the last 8 years.

The info I am basing for the power draw is putting the One X around 100-180Watts.

Besides, I still don't see how a link between this info and specifically the Scarlet was made.

@boxrekt:

Reminder:

X1X has Variable Shading Rate like feature. https://www.eurogamer.net/articles/digitalfoundry-2017-the-scorpio-engine-in-depth

Andrew Goossen tells us that the GPU supports extensions that allow depth and ID buffers to be efficiently rendered at full native resolution, while colour buffers can be rendered at half resolution with full pixel shader efficiency. Based on conversations last year with Mark Cerny, there is some commonality in approach here with some of the aspects of PlayStation 4 Pro's design, but we can expect some variation in customisations - despite both working with AMD, we're reliably informed that neither Sony or Microsoft are at all aware of each other's designs before they are publicly unveiled.

Depth buffer deals with geometry render buffer

https://developer.nvidia.com/vrworks/graphics/variablerateshading

PC "NAVI 10" doesn't have Variable Shading Rate feature.

PC "NAVI 10" doesn't have hardware accelerated ray-tracing feature.

Flute's 1.6 ghz with 3.2 Ghz boost seems to be ultra-mobile configuration like my Ryzen 5 2500U based 2-in-1 tablet/laptop.

.

Don't even care what Scarlet is packing, my Xbox One X will do nicely and I'll probably jump into Scarlet 1 year later or so.

The next batch of consoles supposed to be more responsive and games take noticeably less time to load or have no loading.

@Pedro: Even so, I want to get out as much as my Xbox One X can cause I paid top cash for it and worth every penny too. And by then, Microsoft should have decent good games for Scarlet 1 year later. Gotta get my money's worth.

Fair enough. We would see next year what they have to offer in general. If they kept to their promise you wouldn't miss out on any games. :)

I doubt the console's will differ much when it comes to the CPU which will be a mobile Zen 2 chip... The GPU and price of the console is where the real competition will be, which company will spend more and charge more for a better console and which will be more conservative with GPU and in turn have a lower priced console.

Its a 200w limit on consoles by law.

Its why PS4 and X1 used 7700/7800 sub 150w GPU's rather than the 200w 7950's that where available.... Also the same reason X1X went with Polaris with a 150w GPU rather than Vega which was a 210 GPU.

Navi XT is a 225w GPU... They will be lucky to get Navi 5700 Pro which is a 180w GPU in addition to a Zen 2 Mobile chip and ray tracing cores to abide by EU power usage laws : https://www.playstation.com/en-gb/legal/energy-efficiency/

Wrong

5700 XT draws 225w when the clock rates exceed 2000 MHz

If you cap the boost clocks to 1800 MHz, the power raw is around 150w

The knee is at 1800 MHz, and surprise surprise, Gonzalo and Flute are also rumored to be 1800 MHz.

Coincedence ?

But that graph shows around 210W for 1.8ghz?

2ghz at 275W

At the wall, meaning the whole PC. Testing was done with a 3700X underclocked to 3.2 Ghz btw.

GPU alone would be around 150w

One problem, X1X has automatic under-voltage profiler per silicon quality VRM design.

Your arguments are valid for PS4 vs HD 7870 and PS4 Pro vs RX-480 results, but not for X1X.

Under-voltage RX-5700 XT resulted in 132 watts power consumption.

Scarlet is being designed in similar pattern to Scorpio e.g. bus width, APU chip area size, better silicon maturity.

-------------

7950 has support for 1/3 rate 64bit FP, hence it's extra silicon with double width register memory storage. Game console GPUs wouldn't have workstation features.

7950's power consumption can be reduced via 700Mhz clock speed and modest under voltage.

I mod'ed near reference Sapphire 7950 810Mhz into Sapphire 7950 900 Mhz with it's existing VRMs. 7950's VRMs are over built for overclocking e.g. 950Mhz OC.

AMD's 7950 Boost Edition firmware update has 850Mhz with 925 Mhz boost mode.

The trouble with undervolting is that its the same as overclocking and on the extreme ends of the spectrums like the one you posted require binning as its not guaranteed to hit the same frequencies and be as reliable with every chip which will cost significant time and money in development.

Not to mention that WE have no idea what impact AMD implementation of ray tracing cores will have on power draw and price of the chips to manufacture.

This thread has been ron'd :)

And besides isn't the name flute trademarked by Sony as its name for its 2018 E3 conference LOL.

@rdnav2 said:

@Grey_Eyed_Elf said:

I doubt the console's will differ much when it comes to the CPU which will be a mobile Zen 2 chip... The GPU and price of the console is where the real competition will be, which company will spend more and charge more for a better console and which will be more conservative with GPU and in turn have a lower priced console.

@goldenelementxl said:

@Pedro: The 3700X is a 65W tdp chip with a 3.6 base and a 4.4 max boost. I thought we were getting a cut down version of that in the next consoles, but not this cut down. The Xbox One X tdp is around 245w. The jag CPU is clocked at 2.3 and is far less efficient than Zen 2. Why not keep the base closer to 2.3 next gen? 1.6 just seems nuts to me.

Its a 200w limit on consoles by law.

Its why PS4 and X1 used 7700/7800 sub 150w GPU's rather than the 200w 7950's that where available.... Also the same reason X1X went with Polaris with a 150w GPU rather than Vega which was a 210 GPU.

Navi XT is a 225w GPU... They will be lucky to get Navi 5700 Pro which is a 180w GPU in addition to a Zen 2 Mobile chip and ray tracing cores to abide by EU power usage laws : https://www.playstation.com/en-gb/legal/energy-efficiency/

Wrong

5700 XT draws 225w when the clock rates exceed 2000 MHz

If you cap the boost clocks to 1800 MHz, the power raw is around 150w

The knee is at 1800 MHz, and surprise surprise, Gonzalo and Flute are also rumored to be 1800 MHz.

Coincedence ?

Surprise Surprise Gonzalo and Flute both have Navi 10 "Lite" as the GPU codename.

• 5700 = Navi 10 XL
• 5700 XT = Navi 10 XT

Navi XT is a 225w GPU... Its not wrong, its "wrong" when you try to artificially cap the cards performance like you with 1800MHz which I understand why you did but the XT is not what its those consoles based on the fact that they will use Ray Tracing cores and have a codename of Lite.

Also where did you get 150w as that chart has a 200w+ usage from the wall?...

@boxrekt:

From what i read in resetera Flute is not actually for Scarlet but for PS5.

Well from what i have read they got this all wrong so did TC,Flute is for PS5 apparently the 1,6ghz speed would probably be for BC with the PS4 without having problems.

This explain why boost speed is 3.2ghz far from 1.6.

Is also worth pointing out that the all 3 code names for this chip Ariel,Gonzalo and Flute are characters in William Shakespeares' play,2 from The Tempest 1 from So A Midsummer nigth dream.

So if this is for a console is all PS5 or all Scarlet.

But who knows i would not be surprise to see another console with gimped CPU clock speed.

I was just thinking, it is July 2019.

There's no Firestrike score for whatever fictional SoCs you're talking about.

3700x is suppose to be 65w however when under full load its more like 90w. So the fact that these console cpu's will have a 1.6ghz base clock and with full load usage be well under 3.2 ghz will be no surprise. Since they have to keep power draw and heat output under control.

At the same time these cpus should be at the very least 5x faster clock per clock than jaguar based cpus. So with the push of 2k,4k+ resolutions cpu clockrate wont be as important at the 30 fps and 60 fps caps

lol correct yourself buddy.

I didn't get anything wrong!

I directly linked the information shown in this thread. I didn't add anything or take anything away from what was provided.

You on the other hand, where is your link that you're using to dispute this information?

1. MS has customized automatic undervoltage profiler per silicon quality VRM design with X1X, hence voltage curve is tighter with MS's X1X.

2. Scarlet's devopment timeline is similar to RX-480's release to Scorpio's release gap and next year's 7nm+ UV has 15 percent power consumption reduction.

"NAVI 10" is missing Microsoft's "hardware acclerated raytracing" declaration.

L3 cache size is halved ie. 32 MB to 16 MB.

1.6 Ghz clock cycle timing target seems to be PS4's 1.6 Ghz CPU.

Unlike PS4's APIs and part-time boost mode, X1X's boost mode is full-time i.e. DirectX 11.X and DirectX 12.X APIs was designed to scale with hardware power.

PS; Current Ryzen mobile CPUs has half L3 cache when compared to non-APU counterparts.

At the same clock speed between Zen v2 vs Jaguar

• Vector math compute power is 4X scaled
• Integer math compute power is 4X scaled

Comparsion between Zen v2 at 3.2 Ghz vs Jaguar at 1.6Ghz

• Vector math compute power is 8X scaled
• Integer math compute power is 8X scaled

-----------

1800 Mhz GPU with 36 CU yields 8.3 TFLOPS which is nearly 2X improvement over PS4 Pro's GPU

PS4 Pro's GPU 911 Mhz to 1800Mhz would be nearly 2X raw GPU improvements i.e.

• 2X scaling quad prim trinagle units
• L2 cache 2X faster per CU
• L1 cache 2X faster per CU

At the same clock speed, NAVI has doubled the wdith with the following hardware

• NAVI 10 includes 2X prim triangle inputs over Polaris/Vega's four prim tringle inputs. Geometry-raster power can be 2X to 4X range improvements
• Texture filiter processing yields 2X scaling, hence the need for 2X scaling with texture read capability from cache and memory.
• PS4 Pro GPU's 2MB L2 cache storage is 2X with NAVI 10's 4MB L2 cache. Under NAVI, ROPS connected to L2 cache for tile cache rendering.
• L0 cache is 2X scaled wdith.
• Local Data Store is 2X scaled wdith
• DCC (delta color compression) applied for all graphics pipeline stages and DCC processing capability scaled by 2X
• 256 bit GDDR6-14000 is 2X scaled from PS4 Pro's 256 bit GDDR5-7000

NAVI 10 36 CU at 1800Mhz vs PS4 pro GPU at 911 Mhz

• Texture filter processing is 4X scale
• Prim triangle processing is up to 4X scale
• L0 cache is 4X scale
• Local Data Store is 4X scale

Sony's PS4 process node improvement from 28 nm to 16 nm is 2.3X GPU TFLOPS improvement with PS4 Pro.

NAVI 10 with 36 CU at 1800Mhz faster than PC's RX-5700 1625 Mhz game boost clock speed.

There's a reason why AMD restricted RX-5700's clock speed from reaching RX 5700 XT's clock speed i.e. raster power would be similar when RX 5700 has 1800Mhz against RX-5700 XT's 1750 Mhz game boost clock speed.

@boxrekt:

https://www.tomshardware.com/news/amd-navi-10-navi-12-navi-14-navi-21,39684.html

"NAVI 10 Lite" is mentioned in Linux x86 PC driver.

But can you shoot the ROPS?

Only if they try to rop you.

All 3 of those are different versions of the same console.

Gonzalo is rumored to be PS5 so I take it these are different versions of the PS5 that Sony will choose from for their final Hardware of choice depending on which of those will be most reliable, efficient and powerful while also being able to keep costs down.

A lot of excitement here why don't we just wait eh?

No one has shown how this info is linked to Project Scarlet or the PS5.

https://www.techpowerup.com/review/wolfenstein-youngblood-benchmark-test-performance/

When new hardware features are used, but variable rate shading is disabled in benchmarks.

Both RX-5700 XT (NAVI 10) and RTX 2070 (TU106) delivered similar results.

Both RX-5700 XT and RTX 2070 has similar TFLOPS, raster and geometry power.

Hardware Async compute and Rapid Pack Math FP16 features with Turing and Vega II/Navi 10 shows GTX 1080 Ti's aging design.

Polaris and Pascal GPUs will age rapidly when new hardware features are used.

Variable Rate Shading feature yields 15 percent frame rate boost and it's better than DLSS.

--------

IF RX-5800 XT has 3 shader engines (six prim triangle units, 96 ROPS, 60 CU, 6 MB L2 cache) and 384 bit bus GDDR6 would be about RTX 2080 Ti level