Glad it's more official now that someone from DF has said it.
Called it way early when I said, devs will work to a known spec and won't be relaying on "boost"
Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]
- Thread starter Proelite
- Start date
- Status
Glad it's more official now that someone from DF has said it.
Called it way early when I said, devs will work to a known spec and won't be relaying on "boost"
Glad it's more official now that someone from DF has said it.
Yes, he said it before though, think it was yesterday, and today on this forum. Its what i was trying to say but he does it so much better
Now i'm not sure how much sarcasm there is in the sentence.Thats.... a good thing
But i give an example: The realtime GI stuff i'm currently working on can actually not saturate the Vega GPU i'm using. The test scene is quite small, but to compensate i set all settings to 'ultra'. It takes 2ms which is fine. Then i look at GPU clocks and i see it keeps running at 166 MHz and is totally bored.
The example is not very practical - in a game this task would run async with other work. But there are situations where no max clock is necessary, similar to how it rarely happens all CPU cores run at 100%.
Why not making the other part faster if this happens? It's no bad thing to have dynamic clocks - it's an additional option.
I don't know how it works! Factor it in anyway!
The thing is, BCPack IS factored in. We know the compression chip maxes its output at 6GB/s. The PS5 compression chip output maxes at 22GB/s. Both are best case figures. PS5's RAW throughput is near the XSX's best case output. There is no chance that BCPack provides an advantage that is being missed.
There's no indication at this point the PS5 can run with SMT disabled.
That's not what it was advertised as.
Which we've been told is very rare.
Dictator is talking out his ass. This is explicitly the opposite of what Cerny says happens. DF are not an official source when they are going on about their pet theories.
Very much of it so
You never know, we don't have all the technical info and plans yet, from either i think.
True we could say there are no dynamic clocks, if its that rare its nothing to worry about really.
Source? your ass?
That's what I mean by the SSD situation being the reverse of the DRAM situation for the PS5 and Xbox SX. The Xbox has a raw throughput advantage with a possible caveat in capacity/arrangement in DRAM, while the PS5 has that for its SSD and compressor.
We have all we need: The theoretical maximums. He is positing a situation where BCPack would have to magically exceed the capabilities of the hardware to confer his dreamed of advantage.
Mark Cerny. He gave a talk this week.
What possible caveat for the PS5 SSD? Other than the total capacity being smaller it is straight up faster in every way we know of.
There seems to be a general confusion on here about not being able to distinguish between clock speed and current. Not saying it’s you, just speaking in general.
The load you put in at a given frequency will determine the voltage needed and thus the power draw.
Let’s say a cpu doing 3ghz pulling 20amps at 1v is generating 20w of power draw. 20w is it’s max budget. We won’t go into vrm complications on here.
Lets drop that down to 10amps. There no reason for the frequency to drop as it’s below the max power budget. So at 10 amp load the cpu is happily signing along at 3ghz (max set) drawing 10w.
Now a 25amp load comes in. Because you’re power throttling, your cpu frequency will drop as needed until ”voltage x amps = max power draw.” That’s the “boost” part.
In this situation, it’s actually reverse boost meaning its controlling frequency drop from best case scenario based on load coming in and the power budget allocated.
Yes, but nothing cerney said conradicts what DF/Dictator explained.
So is it correct to say the following?
In this situation the CPU and GPU are part of 1 single unit that has an upper power draw limit. So if the GPU portions are mostly maxed out drawing the majority of the power, you have less power to allocate to the CPU.
It would be useful if there was a chart or appendix of empirical measurements of CPU and GPU instruction mixes and what impact it has on the timings. Since the frequencies will vary, the latencies of the instructions would vary too.
Maybe even a listing of instructions or instruction mixes and their power draw impacts.
AbsoluteBeginner
Regular
So I just thought I should look into how well does RDNA scale with clocks and here is actually interesting benchmark on YT with 5700 (36CU) and 5700 (XT).
In this case, 5700 is clocked at 2150MHz (9.9TF) and 5700XT is clocked at 1750MHz (8.9TF).
Results are surprising because it seems XT pulls ahead most of the time, and sometimes not even by a little, even though it has 400MHz and 1TF deficit. Not sure what to make of this really...
In this case, 5700 is clocked at 2150MHz (9.9TF) and 5700XT is clocked at 1750MHz (8.9TF).
Results are surprising because it seems XT pulls ahead most of the time, and sometimes not even by a little, even though it has 400MHz and 1TF deficit. Not sure what to make of this really...
The caveat is primarily the smaller capacity, and if the BCPack algorithm has the stated advantage on game content there is an additional capacity disparity. On top of that, if the BCPack algorithm has that level of advantage, it would increase the mix of higher-compression assets being read by the SSD versus one using zlib or Kraken.
That average also depends on how much the Series X can apply BCPack versus zlib, which is something I likened to a slower/faster divide in the storage.
In both cases, I think the signs point to the platform with the raw capability having an edge despite having some disadvantages.
It's ambiguous at best. Cerny said the gpu and cpu would run at the max clock, or close to, the majority of the time. Saying dev will have to chose between cpu power and gpu power is in conflict with the presentation, unless the majority of the time means to include optimizations from devs to micro-manage every last percentage.
Again, we are all talking out of our asses since we don't have reasonable real world examples of how often and by how much the frequency drops, and whether the devs manage this or not. HOWEVER the presentation wording suggests against this with "majority of the time", and "very small drop since 2% drop is 10% less power".
Other than Cerny saying the GPU and CPU both run at their max clocks almost always? Dictator is claiming max clocks are an exclusive choice, one of the other. That's not how it works at all.
- Status
Similar threads
- Locked
