Seems multiplats will be the same

Initiating damage control.

Post 1.

Post 2.

Post 3.

Damage control complete. Disaster averted.

It isn't from Sony or Cerny but from your very own article, this is what they have to say:

@tormentos: Dude chill , I just asked you a question , why you get over defensive ?👀

@pc_rocks:

I didnt say it was from Cerny in fact claimed they rounded the numbers as possible answer to the whole 10.3TF because like him i have see that 10.3 shit elsewhere as well.

But again from 10.28 to 10.30 i dont see a single frame more of performance gain.

10.28 or 10.3 is not in question. I specifically asked where did Sony/Cerny claimed that PS5 is 10.3 (or for that matter 9.2) which the TC claimed.

Of course most multiplats will look and perform the same. We are talking about 2 systems that are many times more capable than what's available now, with a difference between them smaller than 20%.

Even if there was a huge hardware difference, most devs would not even know how to take advantage of the new hardware for the first 2 or 3 years.

Very few if any devs are going to take the time to truly optimize games on the XSX over the PS5 especially since the games have to run on the XSS. Most multiplats will look and play almost identically with perhaps some better ray tracing or special effects. Later in the generation it may be evident in exclusive games but time will tell.

@tormentos: Still waiting for you to explain how variable CPU and GPU speeds will manage to run at max clock speeds like XSX as you consistently claim every time you bring this up.

Just the 23 Developers within MS Game Studios I guess :D

The CPU and GPU have variable speeds because SmartShift adjusts power based on a power budget. The shift in power can happen multiple times in a single frame and depending on the frame's complexity and can affect the frequency of the GPU of CPU. From what I learned about all this, the power budget for the chip is enough for both the CPU and GPU to run at 100% frequency at the same time.

The shift happens when you have one component needing more power (AVX CPU instructions for example) when the GPU doesn't need as much. Instead of feeding both components and the GPU and having the GPU idle with that power, the power is shifted to the CPU instead of taken from the budget. This keeps thermals and power consumption down.

Also, when the shift happens the CPU or GPU is undervolted, not throttled down. Undervolting AMD CPUs and GPUs don't translate to linear decreases in frequency. So even when it shifts some power away, the result is literally a few MHz for a millisecond or two.

Hopefully that clears things up.

tc never confirmed his claim with sources, yet this thread is still open. That tells you a lot about how desperate SW needs posters and traffic.

@Zero_epyon: Exactly. Per Cerny at the road to PS5, the GOU can manage those clock speeds as long as there is CPU overhead. They will not manage Max clocks together. Also, given the fact now we have reports that PS5 needs to adjust fan speeds based on user feedback because the system runs hot.

Don’t be shocked if the clock speeds drop because the system can’t manage to contain the heat.

I think you might have misunderstood my post a bit. The GPU and CPU can run at max at the same time. This is what Cerny called the "worst case" scenario when rendering a frame because that's when a frame needs both CPU and GPU running at max frequencies to maintain performance targets. This means that these frames draw the most power and can produce the most heat. But they don't anticipate this happening all the time which is when the shifting will happen.

As for your point with the fans, I think Sony is preparing for both scenarios. There might be a game that doesn't produce a lot of heat and they can tune the fans to be quieter where it was probably spinning higher than needed.

There can also be a game where the system doesn't spin up the fans enough and they'll have to adjust it. I think that building this into their system is pretty damn awesome because we won't be stuck with a jet engine if it thinks the game will be a heat maker when it isn't.

Last point. Clock speeds are not tied to thermals. That's the point of smartshift. Clock speeds are affected by power needs, not how hot the chip is. The thermal management system in the PS5 just cares about the chip temp and fan speed, not about throttling the CPU/GPU.

@Zero_epyon: This exactly

@Zero_epyon: https://youtu.be/KfM_nTTxftE

5min mark on.

Won’t happen.

Also, PS5 bottleneck is its bandwidth. Those high clocks are not helping but hurting the GPU because it’s not gonna put out performance need of it had more CU cores.

Richard made that claim based on comments he recieved from devs saying they used profiles that override smart shift and shifted most of the power to the GPU so that the GPU clock stayed at 2.23. We learned shortly after that this is a feature only available to devs on dev kits. Smartshift will take unused GPU power away from the GPU and distribute it to the CPU if it needs it.

EDIT: Also note that in the same video, Richard confirms from Cerney that the CPU and GPU will, for most of the time, run at their typical clock speeds and reduction of speeds will be minimal.

As for bandwidth bottleneck, where are you getting that from?

@Zero_epyon: out of his a$$, just like everything else Lems pull about PS5 specs.

show us the frame rate graphs!

You spewed a lot of nonsense, and none of them addressed my post. TBH tormy, you calling out some assumed leaning, doesn't do much for your rep. You spew verbal diarrhea on the regular.

I'll simply touch on the fact that I am supposedly running DC for the Xbox, when the TC didn't substantiate their claim... But you do you.

Did you prove your OP yet?

@Zero_epyon: You explain things well. Some people read what they want to read though.

In The Road to PS5, Cerny said they had to cap the frequency to 2.23GHz (presumably, it could have gone higher) and that he expects the frequency to be most of the time at the cap or close to it. He showed 10.3 TFLOPs on the screen, claiming this was the max theoretical number.

@Zero_epyon: It’s all in the video. 🤓

@lundy86_4: where did you get the notion that the settings would be different. Resolution and framerate would change before settings.Have any proof to to dispute that trend? Name a console game that announces 4K60 High, Medium, etc? You can't. Games this gen didn't do it, games next gen won't do it. Post something serious or be done. Thanks.

You made the initial claim, thus you are obligated to provide the proof. Are you all there?

Prove your claim, or just f*ck off. Either is fine with me.

@lundy86_4: you sound delusional . No game ever announces "4K60, High settings". Not on console. Gtfo off my thread bruh. You want me to prove something that's been a trend for years, lmao. Deal with the fact that MS over hyped their console.

I sound dumb? I'm not the one who can't prove my own claim...

@lundy86_4: Not pulling up a link for common sense. The settings for Pro and X were the same and the only differences were framerate and resolution. With a much smaller difference here, you shouldn't expect high settings for one and medium settings for the other. Get some common sense. I'm done with you now.

I see you edited, so allow me to add:

Prove your point. Provide evidence. Do the bare minimum to back up your argument. Stop side-stepping the point of contention. Do the very least of what is required in a debate, which is providing your evidence.

@lundy86_4: I proved my claim. All those games run 4k60. What's next, you're going to ask the color temperature? Gtfo.

Above^^

@lundy86_4: learn to read and click the links. Not my problem you can't figure it out. Sorry.

You said they run the same. There's no evidence of that. 4K60 with mixed High settings is not the same as 4K60 with mixed High-medium-Low. So, are they running the same or no? Where's your evidence?

Yeah in the video they managed to increase the ram speeds by 9% and only managed to squeeze out an extra 1% on average. The Series X has 10GB of memory that's about 20% faster than PS5's 16 GB of memory thanks to a wider bus. So if the same memory performance holds in the next gen consoles, the PS5 would only perform about 2-3% better if it had the extra memory speed. Not really much of a bottleneck.

Thanks. Unfortunately yes.

@Zero_epyon: Almost yes. See, the higher they clock the CU’s the less performance they get. Made worse by the fact they cannot increase the bandwidth.

XSX has such a performance advantage it’s almost comical. Once developers start focusing on the new tech, XSX will establish itself as they premiere console of the generation. At least for the first 3/4 years and the Series Elite launches.

Yep, that's why I said exclusive games. Although not all 23 studios will put games out that look better than anything else but a few will probably do their damned best. If they don't downgrade Hellblade 2 that is looking incredible and easily better than anything else I have seen on console (and by no small margin.. that game looks incredible). The question is, how long will it take until we see some of these games? 2-4 years? At that point the question is whether or not one or both will have mid gen refreshes which would kind of detract from the point.
I am interested to see how things play out, it just sucks we can't know now instead of having to wait to see for ourselves.

That might be true in some cases. But take a look at this:

https://www.pcgamesn.com/amd/radeon-rx-5700-unlock-overclock-undervolt

Essentially, when you remove the artificial limits of the RX 5700 and clock it higher, it can perform as well if not better than the 5700 XT at stock speeds with the extra CUs. More examples here:

AMD knows this. That's why AMD had the software limits in place so the 5700 and the 5700 XT didn't perform identically while both were on the market at different prices. Sony and MS most likely knew this as well but they obviously chose different strategies. But the above demonstrate that a higher clocked lower CU count GPU can rival a lower clocked higher CU count card of the same architecture.

DF did a quick test but I doubt they went as far as using the MorePowerTool. Instead, they downclocked both cards and did the comparisons there. Not really what's happening between the two consoles.

The memory bandwidth is a non issue really. ~2% difference in performance is negligible and couple that with the fact that there's only 10GB of memory that fast on the Series X while the other 6 are slower than the PS5. Split memory architectures are always less than ideal, but we'll see how devs work around that.

So don't expect the differences to be night and day. I predict that Xbox will win a few and PS5 will win a few, but in either case, the differences will be minor. Oh and don't forget that the tech in the Series X is the exact same tech in the PS5. Except one is clocked higher.

@SecretPolice said:

That's funny, i'm gonna steal this for future use LOL.

@Zero_epyon: Yep. I have seen that video many times. It’s all Torm ever post during his long essay as he has a total meltdown.

Here is the problem with your argument and video that for some reason you are not understanding. Whenever people point this out to your friend who also uses those videos, he always seems to vanish......

You ready? DF ran the test using a CPU that is near identical to what will be in the next generation consoles. That is key.

So when I say it’s all in the video. I mean it’s all in the video. If you would watch from 5 mins on you would see OC that GPU is gonna get them less performance. They have a massive CU and memory bandwidth disadvantage. Now, will games look noticeably better on XSX? Maybe. Will games have better performance? Definitely.

The card inside the PS5 is a 2060 Super (or 2070 take your pick).

The Card inside XSX is a 2080 Super (Its actually better, but let’s just say 2080 Super).

Now go look at performance between those card and downplay the power difference.

@tormentos:

its a huge gap,i wanna see it

:P

This is factually false or else there won't be a need for SmartShift. People saying that should know that SmartShift isn't new and already been in used on Laptops. Why would you shift the power and to where if both can run at the max frequency at the same time. Put it another way, why would a component need more power when it's already running at full power. What you're describing is an oxymoron.

Lastly, if none of the factual arguments sway you Sony people's minds then we already have devs saying they are throttling CPU to make GPU run at full clocks. No matter how many mental gymnastics Cerny or cows do, the fact remains the same that the CPU and GPU can't run at full clocks at the same time.

As for these were only available on Dev Kits excuse that Cerny provided back then, why did he refused to answer the question on base clocks when DF asked him point blank. He could have said it simply that 2.23 and 3.5GHz are base clocks.

Not my point. I said where did Sony/Cerny claimed PS5 to be 10.3TFLOPs machine. They called it the peak performance not the sustained or base performance. What's the number when GPU or CPU downclocks due to power budget distribution according to workloads?

Again factually false and cherry picking workloads that don't stress CUs. If higher clocks are a substitute of more SM/CUs as Cerny claimed then none of the top GPUs have more cores and less clocks which is generally the case in PCs. Apart from pixel rate none of the graphics workloads scale well with clocks over cores. It's baffling that suddenly Cerny and cows know more about GPUs then Nvidia + AMD combined with decades of designing GPUs.

Lighting, RT, Geometry/Tessellation, Geometry culling all scale well on cores, especially RT which is also significantly heavy on bandwidth. Another disadvantage of higher clocks over core count is latency/access time for memory. In other words higher clocks over more cores needs even higher memory bandwidth which PS5 already has less.

First, SmartShift is literally new and only one laptop currently has it and there won't be more until next year.

Second, both CPU and GPU can absolutely be boosted at the same time. I don't know where you guys are getting that it can't. How do I know? Well, because AMD says so:

https://www.amd.com/en/technologies/smartshift

A new interface within AMD Radeon Software Adrenalin 2020 Edition makes it easy to see how power is being shifted to the CPU and GPU.​

Unlike other implementations, AMD SmartShift can boost both components during the same workload.

The point of smartshift to effectively shift power between CPU and GPU when necessary. If the CPU is working on power intensive operations, instead of throttling the CPU to keep it cool, it will get the power that would have gone to the GPU if the GPU is idle. And vice versa. But for normal operations where both GPU and CPU are being worked, they will be power boosted at the same time.

Dev kits let you override the shifter but we all know in production SmartShit uses machine learning to auto adjust. Perhaps devs are given that ability to run experiments. I don't know. I also don't know why Cerny didn't want to mention base clock, but it's probably because it's a moot point since the clocks stay boosted until something comes along that might demand too much from either component.

Smartshift is used to balance the performance on a set power budget. It is only present because neither component is being boosted at the same time. At the moment, the effectiveness of the tech has not been validated and the non boosted performance of the PS5 has not been disclosed. Sony was legally obligated to state variable frequency and max frequency because the frequency is not always running at max despite claims .

Can you elaborate on why the CPU is key to the comparison? The CPU isn't a factor as the scene rendered in the test is GPU bound as Richard points out. In the benchmarks in the video, the same CPU was used for all comparisons at varying resolutions.

You keep saying the same things despite the contrary existing in the video. And now you're saying the Series X is a 2080 super? Yeah can't really take this conversation seriously after that.

Sure if you didn't like the games in the first video and the ones tested here: https://www.pcgamesn.com/amd/radeon-rx-5700-unlock-overclock-undervolt then check out the one above that uses games like RDR2, FH4, Tomb Raider, and Witcher 3.

An overclocked morepowertool edited 5700 can perform as well or perform better than a stock 5700 XT with extra CU counts. This has been common knowledge since last year amongst PC enthusiasts. Also, as Richard from DF states, oveclocking memory on AMD cards doesn't always translate into a significant boost to performance. In his test, he overclocked the memory by 9% but got back a 1% difference in performance.

If games are designed with DX api ( PS5 Oberon ) is going to have a hard time , Oberon doesn't support DX api , Yeah I know OpenGL is just as good but considering some developers are just money grabbers they won't put much effort to create equal versions and if PS5 presents some difficulties even less effort.