Right, but we can make some well educated guesses since we know the physical dimensions of the SeX chip and the performance figures we get from that lines up with what MS has been saying about their performance targets.WHAT WE DON'T KNOW AT ALL BUT PRETEND TO:
- How strong these consoles are
No.I wonder why Jim Ryan failed to mention "hardware ray tracing" when listing PS5 features? Forgot or Freudian slip. Curious.🤔 https://youtu.be/E429XW4isFs
FlashArray//X moves beyond the legacy SSD architectures that are architected to make flash pretend to be a hard
disk. Instead, DirectFlash within Purity speaks directly to raw NAND with a super-efficient NVMe protocol and leverages
NVMe-oF for even faster network speeds between the array and application servers. DirectFlash is implemented
in four components:
Purity Reduce: The FlashArray leverages five forms of inline and post-process data reduction, including compression
and deduplication. Data reduction is always-on and operates at a variable block size, enabling effective reduction across
mixed workloads without tuning. Because different kinds of data compress differently, it applies multiple compression
algorithms over time and uses machine learning to identify the best compression for your workloads. Data reduction
averages an industry-leading 5:1 with a total efficiency of 10:1 (including thin provisioning).
In the experiments we first run microbenchmarks that issue random read requests of different sizes (8 KB, 16 KB, and 64 KB) and 8 MB write requests to each of the three devices. For SDF we use 44 threads—one for each channel— to exploit its hardware parallelism. For the other two devices, only one thread is used because they expose only one channel, and the thread issues asynchronous requests. For SDF all requests are synchronously issued and the benchmarks issue requests as rapidly as possible to keep all channels busy. The results are reported in Table 4. Because SDF does not allow writes smaller than 8 MB we use only 8 MB writes in the test.
From the table we can see that SDF enables throughput close to the architectural limits: for read its 8 MB-request throughout of 1.59 GB/s is 99% of the bandwidth limit imposed by the PCIe interface, and for write, 0.96 GB/s throughput is 94% of the flash's raw bandwidth. Even for reads with requests as small as 8 KB the throughout is 76% or more of the PCIe bandwidth limit.
I consider this as the good last piece of information on PS5 with something confirming the custom SSD. Using the search "the built for purpose SSD storage" in Google. I find great information, this time linked to custom SSD in a data center using the same name. And some information about complex data compression and compression ratio. Currently, SSDs in datacenter use another name and they said exactly the same thing I said for months.
Here comparing it to an HDD they talk about keeping high performance without the need to duplicate the data and they talk too about adding complex compression they have an efficiency of 5 to 1 but this is not the ratio compression because it involves the deduplication of data but if for example, the compression ratio is 4 to 1 using hardware decompressor like in the Sony patent an SSD with a sequential speed of 3.5 GB/s you can reach a maximum theoretical speed of 14 GB/s. This is whey they show this very high number in Sony patent. The difference you don't need machine learning your workload will be precise with gaming and using hardware decompressor you will have some precise compression algorithms.
And the last secret but I talked about it is data predictability of videogames and like currently on HDD having the data as sequential as possible but because random reads aren't as costly than on HDD and the I/O is parallel you don't need duplication of data. If the parallel I/O can cover some random reads it is ok, you will never have a problem. This is the case here in this example concerning Baidu. The data is not fully contiguous, here they write at minimum 8 MB of data sequentially after they need to do a random read but they have high read efficiency.
Baidu Goes Beyond SSDs – The SSD Guy Blog
thessdguy.com
You can reach nearly theoretical raw bandwidth SSD speed.
They aren't replicating industry grade flash NAS/SAN appliances. Not only do they use specific compression mechanisms but do it for creating hashes to avoid sector re-writes on single drives - it also is block level writes across a drive array. Very different than what would be in a console.
If we want to look at advanced performance on a single flash drive at this level then reference FusionIO. It won't be any where near that but it's a better reference guide than anything Pure is up to.
That said, neither are correct references to what I'd imagine a 500 dollar console will do which will have lower write/rewrite capacity along with IOPs. Let's not use 20k drive array appliances as a guide for 200 dollar SSDs.
I expect it to cost maximum like this SSD 114 dollars at release and probably less than this.
That is a retail price too so the cost to Sony dealing direct with suppliers for a few chips would probably be $50 or less for 15 million units a year.
I speak about bom cost. I expect the SSD to have a max BOM cost as big as this one if they order it for PS5.
You believe Sony will be spending > $100 per console on the SSD chips alone? I seriously doubt that....
I've said before some of the parts BOM estimates (SSD, RAM especially) are being grossly exaggerated IMO.
I speak about the BOM cost of this SSD it was 114 dollars on release at retail , the BOM cost depending of the number of order is probably between 40 and 60 dollars. I expect Sony to spend around 40 to 50 dollars on the SSD.
This is bottom feeder console warrior trash.I wonder why Jim Ryan failed to mention "hardware ray tracing" when listing PS5 features? Forgot or Freudian slip. Curious.🤔 https://youtu.be/E429XW4isFs
Welcome. To address your theories:
1. Yes, the GCN to NAVI conversion theory has already been discussed and factored in. A fair one to say, had insiders not already said that the conversion does not apply on this case and that the 12TF is indeed 12TF NAVI. This also applies towards the XSX.
2. Could be a possibility. Then again, Cerny's own words were that the PS5 does have RT hardware acceleration. This leads me to believe that Sony is possibly working with AMD on their own solution. Will it be as robust as MS and their VRS solution? Time will tell.
Yes, Kleegamefan said it. You can find it yourself in previous OT's. Its been discussed as naseum.Can you link me to the evidence of the "insiders" saying this for #1? Because I haven't seen that and I am not sure how they would have that information if Sony is keeping it quiet. I mean it would actually defeat the purpose of "security" for third parties.
The entire point of this is so leaks would be taken care of.
As for number #2, they both have hardware RT, it's just the matter of the word "dedicated hardware" verses just "hardware".
"There is ray-tracing acceleration in the GPU hardware," he says, "which I believe is the statement that people were looking for."
Yes, Kleegamefan
#2. How anyone can still be pushing this narrative is beyond me. RT accelerating hardware in the GPU.
Maybe. Maybe not. Getting 10-20 more die per wafer could easily be $10-$20 in savings for a $5 more expensive cooler.my point is that whatever savings they might get from a smaller die will be offset by the need to have a more elaborate cooling solution.
This whole dedicated thing is just breeding confusion. There is no such thing.Can you link me to the evidence of the "insiders" saying this for #1? Because I haven't seen that and I am not sure how they would have that information if Sony is keeping it quiet. I mean it would actually defeat the purpose of "security" for third parties.
The entire point of this is so leaks would be taken care of.
As for number #2, they both have hardware RT, it's just the matter of the word "dedicated hardware" verses just "hardware".
One Senjutsu gone, 10 appeared in its place.
And RT-accelerated hardware in the GPU would not be taking TF's away from the rest of the console tasks?This whole dedicated thing is just breeding confusion. There is no such thing.
As far as the GPU is concerned, you either have built-in hardware support for RT or hardware that is capable of running RT in software. Now any GPU can runRT in software, albeit it will cripple the GPU, but its possible. So if these companies are saying we have hardware-based RT, then what they are saying should be taken literally. They have hardware in the GPU (be it APIs that repurpose specific hardware already existing in their GPU for specific RT tasks that wil "accelerate" the process or components built into the GPU like TMUs are built-in for textures or how current-gen GPUs had ACUs).
Nice find via that link!
I consider this as the good last piece of information on PS5 with something confirming the custom SSD. Using the search "the built for purpose SSD storage" in Google. I find great information, this time linked to custom SSD in a data center using the same name. And some information about complex data compression and compression ratio. Currently, SSDs in datacenter use another name and they said exactly the same thing I said for months.
Here comparing it to an HDD they talk about keeping high performance without the need to duplicate the data and they talk too about adding complex compression they have an efficiency of 5 to 1 but this is not the ratio compression because it involves the deduplication of data but if for example, the compression ratio is 4 to 1 using hardware decompressor like in the Sony patent an SSD with a sequential speed of 3.5 GB/s you can reach a maximum theoretical speed of 14 GB/s. This is whey they show this very high number in Sony patent. The difference you don't need machine learning your workload will be precise with gaming and using hardware decompressor you will have some precise compression algorithms.
And the last secret but I talked about it is data predictability of videogames and like currently on HDD having the data as sequential as possible but because random reads aren't as costly than on HDD and the I/O is parallel you don't need duplication of data. If the parallel I/O can cover some random reads it is ok, you will never have a problem. This is the case here in this example concerning Baidu. The data is not fully contiguous, here they write at minimum 8 MB of data sequentially after they need to do a random read but they have high read efficiency.
Baidu Goes Beyond SSDs – The SSD Guy Blog
thessdguy.com
You can reach nearly theoretical raw bandwidth SSD speed.
No.
Someone on reddit posted a rather interesting theory.
The last State of Play was on Tuesday, December 10.
Exactly 4 weeks later, on Tuesday, January 7,the PS5 logo was revealed.
Exactly 4 weeks later after that, on Tuesday, February 4, the PS5 website went up.
Exactly 4 weeks after that is Tuesday, March 3rd... what do you guys think? Does this sort of pattern often happen just by coincidence? Or is March 3rd the reveal date?
Damn Ray Tracerists.No.
We have no idea how AMD's RT solution scales, so this line of questioning just borders on more concern-trolling and Ray Tracerism.
Deja Vu like last gen.
Under expection for PS, and over expection for Xbox.
PlayStation 2's 20th anniversary is on March 4th :)Someone on reddit posted a rather interesting theory.
The last State of Play was on Tuesday, December 10.
Exactly 4 weeks later, on Tuesday, January 7,the PS5 logo was revealed.
Exactly 4 weeks later after that, on Tuesday, February 4, the PS5 website went up.
Exactly 4 weeks after that is Tuesday, March 3rd... what do you guys think? Does this sort of pattern often happen just by coincidence? Or is March 3rd the reveal date?
Are we now at a point where we are saying that xsx has 12tf and specifc hardware for RT, that doesn't even have to draw power from this 12tf?
And ps5 has 9tf and for RT has to draw power from those?
I see it is getting worst and worst for the ps5, and it can only get better for xsx.
Literally wanted to write the same ... LOL
Mid-gen boxes with noticeable ray-tracing performance upgrades would be appealing. The increase in teraflops doesn't have to be double.
Screw it... I'm betting my avatar for the 3rd of March Transistor. I hope we get something that day. A WIRED article or at least a date announcement of the PlayStation Meeting or something. Getting real tired of these rumors popping up every other day, not sure which one to believe. REVEAL THE FULL SPECS ALREADY, SONY.Someone on reddit posted a rather interesting theory.
The last State of Play was on Tuesday, December 10.
Exactly 4 weeks later, on Tuesday, January 7,the PS5 logo was revealed.
Exactly 4 weeks later after that, on Tuesday, February 4, the PS5 website went up.
Exactly 4 weeks after that is Tuesday, March 3rd... what do you guys think? Does this sort of pattern often happen just by coincidence? Or is March 3rd the reveal date?
Or the other way aroundI can't imagine the Melt down and tears if the ps5 is more powerful than the Series X once specs are known and figures are released
XD
The file has tests for ArielB0, OberonA0 and OberonB0, so I doubt it's just a random 5700 being tested. Regarding the intern, his job was to collect and aggregate the data, not to do the testing. Each test has a tester name attached to it (I always remove it for privacy when I take a screenshot), there are at least 12 different testers who worked on these tests, not including the intern.Could also be Sony piggybacking on the most similar chip available at the time to do some BC testing.
It was an intern right? Could also be busy work for them with an old version.
Not saying it's those things, but that's why the context we're lacking matters.
That's not how you build a balanced system. Both console manufacturers are probably trying to achieve peak performance and highest clocks they can for their budget (money, thermal and power draw). Sony and MS engineers aren't trying to hit some random TF number so Era members will be happy, they are trying to build the most performant machine they can with the budget they have.
- 36 cu at 2.0 ghz would be a 150w+ gpu. add 40-50w for the cpu and their apu alone is 200w.
- if their goal was to hit 9.2 tflops, they would get there for less than 100w if they went with 1.35ghz and 54 CUs.
- so to save 50mm2 or $10-15, they decided to take a cool a 200w apu that cant even hit 10 tflops.
its hilarious to me that the team that was supposedly team realistic and team thermals for over a year is now putting all their chips in the oberon 2.0 ghz basket. i can actually see sony going with a 9.2 tflops 1.35ghz 54cu gpu. that to me at least makes some sense even if it doesnt line up with other rumors from within the industry. because even then it would be a 150w apu that would require some kind of crazy cooling solution considering the entire ps4 pro consumed 150w.
with n7p and other arch improvements that result in 20% power savings, i can see 2.0 ghz being somewhat viable at around 120w for the gpu alone. but the same should apply to the 54cu configuration which would be under 80w for the same amount of tflops.
i wouldnt be surprised if the oberon 2.0 ghz tests are for BC with some kind of native boost mode that runs all pro games at native 4k 60 fps while disabling the remaining CUs and using the power left from disabling those CUs into increasing the clocks on the 36CUs. sony was afraid to put boost mode out at launch, but the mere 111 mhz clockspeed boost didnt cause any issues on ps4 games with no pro support. now imagine what they can do with 1,200 mhz with the same amount of CUs.
March 3rd - PS5 BR box art reveal :)Someone on reddit posted a rather interesting theory.
The last State of Play was on Tuesday, December 10.
Exactly 4 weeks later, on Tuesday, January 7,the PS5 logo was revealed.
Exactly 4 weeks later after that, on Tuesday, February 4, the PS5 website went up.
Exactly 4 weeks after that is Tuesday, March 3rd... what do you guys think? Does this sort of pattern often happen just by coincidence? Or is March 3rd the reveal date?
I can't imagine the Melt down and tears if the ps5 is more powerful than the Series X once specs are known and figures are released
XD
I think lesser on this one, as many of us expect it will be weaker than XsX. I will be shocked by massive surprise if other ways.
I just want both of them to be very high and very close.I can't imagine the Melt down and tears if the ps5 is more powerful than the Series X once specs are known and figures are released
XD
I think lesser on this one, way many of us expect it will be weaker than XsX. I will be shocked by massive surprise if other ways.
I can't imagine the Melt down and tears if the ps5 is more powerful than the Series X once specs are known and figures are released
XD
As a Sony fan, I would be really, really disappoint, because it would be that PlayStation 5 is either as expensive as Xbox Series X either Sony take a bigger loss on each console.I can't imagine the Melt down and tears if the ps5 is more powerful than the Series X once specs are known and figures are released
XD