Next Generation Hardware Speculation with a Technical Spin [2018]

Status
Not open for further replies.
I wonder if it would ever make economic sense to have a lower-end SKU to allow for salvaging parts with minor defects and conversely having your flagship SKU fully enabled with all functional units instead of disabling some on every chip to improve yields.

Edit: And you could then also drop SoC/memory clocks as well to create a little more of a performance spread with the lower end SKU and save a little money on power and cooling to better enable the lower cost.

I'm still not sure it makes overall sense, but there may be more merit in the idea than is apparent at first thought.

In theory it could be done but would 2-4 extra cu's justify another SKU for consumers to get confused by and devs to deal with? What would they be called ps5 pro and ps5 80%?

I don't think it's worth it but it's an interesting thought. This kind of practice is suited only to PC imo.
 
Both 360 and PS3 had inferior/superior skus at launch but that it did little market share wise for either manufacturer, you could argue that the pluses for the $599 PS3 weren't worth $100 or that the negatives for the x360 weren't worth it for the savings there but that makes the point. The $599 PS3 represented the most additional non critical gubbins you could add to a box that wouldn't increase costs (a USB card drive and chrome accents) and the x360 cheap sku cut an actually useful feature for bugger all cost saving and quite a bit of consumer pain (not at launch perhaps but as soon as game installs became a thing definitely). I mean even the WiiU dual sku strategy didn't help that devices sales as the base costs of a that console couldn't be offset by not including peripherals such as a charging dock. Didn't MS run a dual SKU strategy for a while with XB1 and XB1 + Kinect for a brief period there also (not a great comparison point even if they did, given that was a reactive strategy when everything else was on fire too)?

If you get into bin sorting or selective feature disablement you have to start by designing everything about the system for the faster more expensive sku such as cooling design and power, this saddles the base unit with unnecessary cost hurting your mass market appeal for a margin harvesting sku you are expecting to make up a minority of your sales. It also leads back to the marketing problem of "cool visuals PS5Pro/XB2X only" disclaimers on ads, the mass market hates options they want to hear that the new cool games box is out and that's basically it.

I believe part of this is that buying a console means betting on a closed software ecosystem for several years, that your platform will get the cool exclusives and offer the best gaming experience for the next 6-8 year console lifecycle. If I have to make a 50/50 bet on two eco systems a lot of folks are going to place the minimum stakes, as who wants to bet 50% more than the next punter and "lose". In an open console eco system where the modern equivalent to a PC98 or MPC standard defines the new baseline multiple skus makes sense as no matter my stakes I "win", I'm much more likely to feel confident that my 50% extra investment will return rewards over this life cycle. Absent that multiple skus offers little to the mass market consumer and an awful lot of manufacturing risk.
 
Both 360 and PS3 had inferior/superior skus at launch but that it did little market share wise for either manufacturer

But neither had any differences in the core APU+DRAM. It was primarily different storage options.

The 360 Arcade and later 360 Slim 4GB were important because they hit key price point, with the Arcade SKU being the first of the PS360 to come under $300 (at $279), the 360 Slim 4GB hit $199. Both sold well. They would have sold better if externally attached general storage had been an option (instead of the overpriced HDD capsule)

We might see a repeat of this. A full launch SKU with HDD+optical, and an Arcade-like SKU with just 250ish GB Flash. Heck, if your internet connection is fast enough, you might not need more storage with FastPlay.

Cheers
 
Did he really? Unless I’m incorrect, Spencer’s hint was just using the next Xbox in the plural form. What exactly did he say that makes you think multiple xbox configs at launch versus over a gen like we have now?
I'm pretty sure he's referring to the imminent next gen launch time frame since that's what people are looking forward to next the most anyway.
 
I'm pretty sure he's referring to the imminent next gen launch time frame since that's what people are looking forward to next the most anyway.

I don't get that at all from the quote:

Phil Spencer said:
The same team that delivered unprecedented performance with Xbox One X is deep into architecting the next Xbox consoles, where we will once again deliver on our committment to set the benchmark for console gaming

Given how much emphasis all the gaming press put on the word 'consoles' I would guess it wasn't an accident. I don't think Microsoft have issued a clarification if it was mis-spoken AFAIK?
 
I don't get that at all from the quote:



Given how much emphasis all the gaming press put on the word 'consoles' I would guess it wasn't an accident. I don't think Microsoft have issued a clarification if it was mis-spoken AFAIK?
He's reading from a teleprompter and there are no issued clarifications. So it's safe to assume that yes there is more than 1 currently in development.

What it means is unknown.
 
He's reading from a teleprompter and there are no issued clarifications. So it's safe to assume that yes there is more than 1 currently in development.

What it means is unknown.
Not sure about multiple simultaneous hardware releases. They're probably working on consoles at 2-3 year cycles. They've been structuring their API to support this.
 
All nice with those 2/3 year cycles but those more-powerfull units wont be used to their max, baseline performance will allways be on the standard models. Thats abit in the direction where pc gaming is, an 1080Ti wont be used to max cause a 1060 also needs to be able to run it.
 
https://www.anandtech.com/show/12621/memory-scaling-zen-vega-apu-2200g-2400g-ryzen/3

Not sure if this is old news now but it’s inetersting. Memory bandwidth on AMD APUs.
The test forced everything to C17 latency. It's a weird decision.

Using automatic timings from the memory specs would provide a similar access time regardless of clocking. Faster memory clocks means it takes more clocks to count the same latency period. That's why 2133 CL11 have the same latency as 3666 at CL18.

If the goal was to normalize the latency they should have normalize it in nanoseconds, not in cycle count.
 
8 threads and 4 big cores seems like both a reasonable power boost to me and more in line with trying to produce an APU for a $399 price point, would be very happy to be proved wrong as I think having a lot of CPU power would open up more space for playing with voxels, AI, procedural generation, etc.

A thing I've been thinking about lately is the potential for "free" performance boosts in the memory controller and the like by using the same hacks that were exposed by the Spectre and Meltdown bugs. IIRC they were largely caused by the memory controller in certain AMD, ARM and most every Intel design of the past decade or so not performing boundary checks when doing speculative loads (recently this behaviour has been shown in speculative stores too) which allowed the attackers to force the unsafe loading of protected memory areas. Would there be any advantage to Sony or MS is specifying this unsafe behaviour for their consoles? As devices with pretty strong protections against arbitrary code being run anyway the odds of the attacks being used successfully are considerably lower than with the PC Spectre/Meltdown attacks which largely relied on adjacent attacker controlled VMs running the attack code to compromise other VMs on the same device (there were browser based JScript attacks but simply disallowing high precision timers and adding jitter closed that vector). Or is it more likely that they would be afraid that in the event of a future attack that allowed arbitrary code to be run Spectre/Meltdown could be used to extract root keys from RAM?
 
8 threads and 4 big cores seems like both a reasonable power boost to me and more in line with trying to produce an APU for a $399 price point, would be very happy to be proved wrong as I think having a lot of CPU power would open up more space for playing with voxels, AI, procedural generation, etc.

A thing I've been thinking about lately is the potential for "free" performance boosts in the memory controller and the like by using the same hacks that were exposed by the Spectre and Meltdown bugs. IIRC they were largely caused by the memory controller in certain AMD, ARM and most every Intel design of the past decade or so not performing boundary checks when doing speculative loads (recently this behaviour has been shown in speculative stores too) which allowed the attackers to force the unsafe loading of protected memory areas. Would there be any advantage to Sony or MS is specifying this unsafe behaviour for their consoles? As devices with pretty strong protections against arbitrary code being run anyway the odds of the attacks being used successfully are considerably lower than with the PC Spectre/Meltdown attacks which largely relied on adjacent attacker controlled VMs running the attack code to compromise other VMs on the same device (there were browser based JScript attacks but simply disallowing high precision timers and adding jitter closed that vector). Or is it more likely that they would be afraid that in the event of a future attack that allowed arbitrary code to be run Spectre/Meltdown could be used to extract root keys from RAM?
I don’t think they’ll do anything less than 8 cores due to their backwards compatibility patents. They talk about slowing down and restricting features on a granular basis to match old architectures, and I doubt they want to attempt that on virtual cores.
 
I'm also fairly certain they will sacrifice whatever they need to for full backwards compatibility with PS4. They would surely want you to keep your PS4 library to dissuade you from going with a competitor.

It's the same reason they don't want cross play now.
 
What sort of cores would provide a generational improvement with 4 cores compared to the 8 jaguars?

8 zen2 (or even 12, from the rumors) at reasonable clocks seems more in line with the jump in performance we want. Very fast cores are inefficient, judging from the TDP of various xeon gold offerings at different clocks.
 
Status
Not open for further replies.
Back
Top