News & Rumors: Xbox One (codename Durango)

[liolio]

A Power7, even stripped of some of the non-console stuff like the XML and I/O features for the original market audience, is going to be a big chip, even with a reduction in eDRAM. iirc a Power7 has 32MB of eDRAM for an 8 core.

On the flip side a 4 core Power7 (16 thread) versus a 3 core Xenon (6 thread) is a total laugher. The Power7 runs circles around the Xenon per core. Heck, I wouldn't be surprised if silent_guy or someone else came in and gave an educated guess that 1 Power7 core (4 threads) is faster than 2 Xenon cores (4 threads) by a wide margin.

That sounds like a reasonable bet (per cycle) for anything but super SIMD friendly workloads.

Pretty much like X87 high end cores power7 cores would fly around Xenon. Damned we forgot a lot of the complains at the beginning of this gen in regard to cryptic single thread performances.

Luckily we still have Joker454 to remind us that some code (single thread obviously) was running fast on the xbox 700Mhz celeron than on Xenon

Still we can't tell for sure, it could be power a2, or a new throughput oriented cpu cores.

I had a long post some pages ago about the odds of a cpu cores feeding two SIMD instead of one. I would like the one Silet_guy or others knowledgeable members to tell us their opinion on the matter.
Lets rephrase it so they can answer "how about a modified power7 including two reworked vmx 128 units?"

 

[Acert93]

Is that not what is going on in the x86 world? Does not Bulldozer offer the ability for a module to do 2x 128bit SSE or 1x 256bit AVX instructions (and iirc each Intel core has a 256bit AVX SIMD, and AVX2 coming with Haswell, that can be packed with multiple SSE, but I may be remembering wrong).

So the idea of a "VMX ver. 2.0" in the Xbox3 where each core has 2x 128bit vector units would seem possible? Or depending on the execution unit strategy maybe even more? Also, didn't the way the VMX was designed in Xenon that the peak was lower than 115GFLOPs because you couldn't be co-issuing across the entire pipeline so the real max was just below 80GFLOPs?

 

[liolio]

Is that not what is going on in the x86 world? Does not Bulldozer offer the ability for a module to do 2x 128bit SSE or 1x 256bit AVX instructions (and iirc each Intel core has a 256bit AVX SIMD, and AVX2 coming with Haswell, that can be packed with multiple SSE, but I may be remembering wrong).

So the idea of a "VMX ver. 2.0" in the Xbox3 where each core has 2x 128bit vector units would seem possible? Or depending on the execution unit strategy maybe even more? Also, didn't the way the VMX was designed in Xenon that the peak was lower than 115GFLOPs because you couldn't be co-issuing across the entire pipeline so the real max was just below 80GFLOPs?

A module is more or less two cores, two scalar pipeline, two SIMD (can act as one).

The idea would be 2 SIMD for one scalar pipeline. I wonder if it would be easier to feed two narrower SIMD than a bigger one. It may prove a bit costlier but could it be more efficient?

With the scalar pipeline way more potent than say in Xenon it's (in my views) a given that a core would match two xenon cores even in the worse case scenario.

As I said some page ago it could be a way to amortize the bigger front end and scalar pipeline on 2 SIMD vs one.
I also wonder if OoO could limited to the scalar pipeline to lessen the cost of its implementation.

 

[JB9861]

http://sillegamer.com/2012/04/06/xbox-720-devkit-specs-detailed-includes-16-core-processor/
Rumor number "over 9000", provided by latest issue of Xbox World magazine:
- 16 "core" IMB CPU [most likely quadcore Power7 with 16threads ]
- Radeon 7xxx series GPU [no mention of dual setup]
- Bluray drive
- Kinec taking up to four "cores"
- they are expecting PS4 to be even more powerfull [nothing concrete]
- 3rd party devs will maybe reveal some nextgen games on this year's E3

Well it will be interesting to see if Sony attempts to keep up. I'd be surprised because MS looks to be pulling out all the stops on several fronts.

 

[french toast]

Rumor number "over 9000", provided by latest issue of Xbox World magazine:
- 16 "core" IMB CPU [most likely quadcore Power7 with 16threads ]
- Radeon 7xxx series GPU [no mention of dual setup]
- Bluray drive
- Kinec taking up to four "cores"
- they are expecting PS4 to be even more powerfull [nothing concrete]
- 3rd party devs will maybe reveal some nextgen games on this year's E3

Thats much more like it!...thats the kind of stuff i was initially hoping to hear..and what i was expecting....
Unless its something like 4x power 7 cores 4x SMT fully OoO..@5GHZ...equals= 16 threads or 'cores'..how would perform if they tried to make every thread efficient?

Unless they have got IBM to design a specialised custom core again thats OoO very small and efficient..with out SMT?..other than that i can't see Microsoft fitting in 16 individual processors..connecting all those together is going to be a difficult to say the least.

Rumours of HD 7000 graphics are nice to hear... it doesn't matter that PS4 will be more powerfull than this..no developer wpould be able to take advantage of it anyhow for the first 3 years..

 

[Acert93]

If the PS4 is really a 2 module (4 integer core) A8-3850 (rumors, unsubstantiated rumors) I don't think it is easily said it is "more powerful" than a 16 "core" Power 7 (4 core, 16 hardware thread) design. In fact I would think 4 discreet cores, each with 4 threads, is going to be a good leap ahead of a 2 module / 4 int core AMD APU as a CPU. Now the GPU pipeline could make things interesting but that really falls back to what MS, theoretically with this rumor, was using in the South Island product line. If it is a 7670, i.e. 6670 like the rumored PS4, that creates a situation where the PS4 may be able to do some things better with some elbow grease as well as snag some low laying graphics fruit; on the reverse I would think general code is going to very quickly run very fast on that Power7 as you have a ton of threads to throw stuff at. But if MS has, say, a Pitcairn class GPU it is a cake walk--a 4 core Power7 with a 7850 class GPU is going to stomp all over a A8-3850 CPU / 6670 (7670) GPU design. We are talking at least 2x the speed which is enough that we could be seeing one title at 720p and the other at 1080p or one at 30Hz and the other at 60Hz or one in 2D and the other with 3D. 10-25% performance difference won't shake the world, especially if the slower model is easier to develop for. If the slower model is also requiring more work that could be a problem.

 

[N2O]

well they said "Discrete sources at GDC also confirmed to the publication that developers expect Sony’s Playstation 4 to be more powerful than the next Xbox"

maybe that's why EA angry about PS4 spec later(not what they expected)
but who knows

 

[Megadrive1988]

A believe a 16-core processor is possible. In the book "The Xbox 360 Uncloaked" the engineers talked about that they could fit 8-16 cores on a single chip for Xbox 360. Of course that never happened but I believe it could happen for Xbox3.

 

[ultragpu]

I find it extremely funny and pointless how the author used the word "confirmed" as if it would make a heap of difference. I mean one can confirm their expectation of something without being actually correct about it.

 

[-tkf-]

So what happened to the heat/power discussion? Everything is possible now?

 

[Rangers]

Here's two alleged rumors from an alleged source

DICE is set to receive a new iteration of Xbox Durango development kits by August.

I'm not sure if the 16-cores Xbox Durango rumor is true or not but I do know Microsoft is going the 'insane' route by developer request

 

[DieH@rd]

Anyone knows how power hungry is quadcore/16threaded Power7 CPU [some server variant]?

 

[Proelite]

Here's two alleged rumors from an alleged source

Source?

 

[french toast]

Source?

Tomato

 

[Proelite]

Tomato

Sauce?

 

[Acert93]

Anyone knows how power hungry is quadcore/16threaded Power7 CPU [some server variant]?

The best thing I can find is 100W-170W from Anand.

We now have a proven platform with excellent RAS features that needs slightly less power now while it provides a decent performance boost. That's excellent, but the Xeon E7 still has a few weakness. One weakness is the relatively high power consumption at idle load. Compared to the high-end Power 7 servers, this kind of power consumption is probably very reasonable. The Power 7 CPUs are in the 100 to 170W TDP range, while the Xeon E7s are in the 95 to 130W TDP range. A quad 3.3GHz Power 755 with (256GB RAM) server consumes 1650W according to IBM (slide 24), while our first measurements show that our 2.4GHz E7-4870 server will consume about 1300W in those circumstances.

Considering that the 3.3GHz Power 7 and 2.4GHz E7-4870 perform at the same level, we'll go out on a limb and assume that the new Xeon wins in the performance/watt race. AMD might take advantage of this "weakness", but availablility of quad 16-core "Bulldozer" servers is still months away and we don't know what the power use will be yet.

The high end would be for an 8 core @ 4.14GHz on 45nm. It is said to be a lot better than the Power6 architecture in performance and power efficency. Ars notes on Power7 versus Power6, "First in the chain is the 32KB L1 data cache, which has seen its latency cut in half, from four cycles in the POWER6 to two cycles in POWER7. Then there's the 256KB L2, the latency of which has dropped from 26 cycles in POWER6 to eight cycles in POWER7—that's quite a reduction, and will help greatly to mitigate the impact of the shared L3's increased latency."

Anyways, moving to 32/28nm, cutting out extra silicon not necessary for a console, going with a lower clocked 4 core (16 thread) variant should get a more reasonable power envelop. But... it is still likely to be much higher than a 2 module / 4 int core / 400 SP AMD APU.

 

[Silent_Buddha]

Here's two alleged rumors from an alleged source

If that second quote is even remotely true (I have my doubts), that would be an indication that at least MS plans to continue the "Hardware as a loss leader" route of console design and sales.

Then again, perhaps their idea of "insane" is different than my idea of "insane."

My idea of insane would be...

[1] 64 programming environment. Please, oh please... That might mean more 64 bit versions of games without the 2 GB VA limit in Windows.

[2] Massive number of cores, ok the rumor has that covered.

[3] Enthusiast class GPU. While it'd certainly be entertaining to see that. I just don't see how that would be even remotely possible in a box meant to sit front and center in a family's home entertainment center. Heat - Noise - Size. Pick 2 to reduce because at least one of those is going to have to be very large for an enthusiast class GPU.

[4] Fast SSD main storage with mandatory installs. Considering the rumor is that MS are including a BRD, that's also going to be a fairly large SSD.

[5] 8-10 GB of memory. Depending on how serious they are about 4k (REALLY doubtful). 4 GB main and 4-6 GB video (if targetting 4k, much lower [2-4 GB] if not). Of course, this also means that [1] above would be mandatory.

I obviously don't see that happening for less than 1K USD without MS eating a significant amount of the cost. So I'm going to guess my idea of insane isn't remotely close to their idea of insane.

Regards,
SB

 

[AlphaWolf]

Considering the opinions of many on this forum any box not including blu-ray, backwards compatibility, launch profitability and a tdp under 200 watts could qualify as insane. So, that's a big window.

 

[Ninjaprime]

The best thing I can find is 100W-170W from Anand.

Top end 8 core at 4.14GHZ is almost 250W TDP.

http://www.realworldtech.com/page.cfm?ArticleID=RWT040511235825&p=2

...Given that the 4GHz POWER7 runs close to 250W, even using IBM’s exotic packaging...

 

[french toast]

Sauce?

Ha ha