• Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
  • We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.

How much money are you willing to pay for a next generation console?

  • Up to $199

    Votes: 33 1.5%
  • Up to $299

    Votes: 48 2.2%
  • Up to $399

    Votes: 318 14.4%
  • Up to $499

    Votes: 1,060 48.0%
  • Up to $599

    Votes: 449 20.3%
  • Up to $699

    Votes: 100 4.5%
  • I will pay anything!

    Votes: 202 9.1%

  • Total voters
    2,210
Status
Not open for further replies.

Pheonix

Banned
Dec 14, 2018
5,990
St Kitts
He would if the initial plan for ps5 truly was 2019 and MS got wind of that. It wouldn't even be obfuscation on Sony's part, shit happened, it got delayed and they had time to work on things. What's more interesting to me is if Sony designed a more powerful console by reengineering an initially 2019 design than MS who always planned to have 2020. That's a bit of egg on your face, IF that's the scenario that took place
That's just sensationalism lol. If you are launching in 2019, then you are most likely building your APU around tech that was available in 2018. There is a ceiling of power that you can hit at a given price range.

Sony and MS releasing in 2020 means just that. They are both basing their hardware off an APU that is made of a CPU and GPU that is present now in 2019 (with some modifications).

I'll say it again, both the PS4/XB1 launched around the same time in 2013. Its no coincidence that both machines are near identical with regards to their architectures. They just differ in one major way that not only resulted in using different types of RAM but also accounted for the difference in performance.
 

Kiekura

Member
Mar 23, 2018
4,043
Whats the point to release 6 AAA exclusives in one year? The there would be timespan where we dont get big exclusives at all and not to mention those games would eat each others sales and would need to complete against third party games too. It isn't happening, even If it was possible.

We will get 2-3 AAA games from Sony each year max.
 

Jaypah

Member
Oct 27, 2017
2,866
Ya, I included physics in with animation (in a game scenario would be different because animation doesn't necessarily rely on the same reactants as physics). We're getting closer to that in real time however. TLOU 2's motion-matching brings us one step closer. Physics and animation will probably be the next big leaps in immersion.

Fair enough. And yes, their motion matching is mind-blowing. Honestly I would have believed that the demo was early next-gen stuff.
 

Deleted member 12635

User requested account closure
Banned
Oct 27, 2017
6,198
Germany
We can test the effect of Variable Rate Shading (VRS) on our own GPU in a not so distant future with 3DMark. The feature is launching 26th August this year.

Wccftech Article

3DMark feature tests are special tests designed to highlight specific techniques, functions or capabilities. The 3DMark VRS feature test is designed to help you compare differences in performance and image quality when using variable-rate shading. The test also offers an interactive mode for experimenting with different VRS settings and exporting frames for comparison.
 

kinjx11

Banned
Jan 24, 2019
299
This guy has seen the Gonzolo score so i would take his word for it. We are getting an 8.5-9 tflops gpu

what doesnt make sense is how the ps5 can be stronger than the xbox. what was Phil going for? 7.5 tflops scarlett? I refuse to believe he went up on stage at E3 2018 claiming he will set the benchmark for next gen consoles? would he really do that with just a 8-9 tflops GPU?

or the safest explanation is that reiner is wrong, and so is colin. people are confused because Sony's devkits have a vega 56 running at 1.8ghz to simulate a 1.8 ghz 8tflops RDNA GPU.

i think AMD is working on a DLSS-like technology for consoles which isn't new at all but with deep learning the IQ can be much sharper
 

sncvsrtoip

Banned
Apr 18, 2019
2,773
firestrikepowersijyq.png

[email protected] restricted 5700xt to 1749mhz (8.95tf) firestrike score 20208 system power draw(peak) 208w
 

dgrdsv

Member
Oct 25, 2017
11,886
This may be a dumb question but why don't devs use FP16?
FP16 precision is too low for the vast majority of graphics processing. Rounding errors accumulate fast and on a big TV screen become very apparent, less so on a smartphone sized displays which is why it's used widely in mobile gaming.
There are still many applications for FP16 though, if not in pixel shaders then in compute and simpler stuff like post processing. It will be used on next gen machines for sure but not to the extent of making them twice faster of course. Just some low percentage improvements on average here and there.
 

Deleted member 12635

User requested account closure
Banned
Oct 27, 2017
6,198
Germany
firestrikepowersijyq.png

[email protected] restricted 5700xt to 1749mhz (8.95tf) firestrike score 20208 system power draw(peak) 208w
People should stop undervolting the GPU and the CPU in these tests and instead decrease the boost clocks for the CPU/GPU. That way silicon lottery doesn't come into play. IMO those tests are not drawing the correct picture.

You can easily underclock a GPU by using MSI Afterburner by moving down the clock stages as a whole iirc. Just click on the left symbol for the clock graph that will open a second window. There select the max clock speed point and press the "Shift" key while you moving the point up or down with the mouse. This way you can move the clock curve easily as a whole.
 
Last edited:

sncvsrtoip

Banned
Apr 18, 2019
2,773
GPU score is 24k which is the same as a RTX 2070 and 2k lower than my RTX2080.

Disappointing but this seems reasonable. i would like to see some actual ingame benchmarks with these specs. would it still come close to outperforming vega 64?
8.95tf navi ? it's little faster or ona pair with rtx2070, and easly eat vega64. Stock 5700xt is on avarage 9.3tf so only 4% faster
 

anexanhume

Member
Oct 25, 2017
12,913
Maryland
A bit off topic but i am really amazed by Samsung achievements with memory:

Samsung Commences Mass Production of First Ever 12Gb LPDDR5 DRAM for Premium Handsets

Too bad nothing happened with that story from few years ago about new low-cost Samsung HBM for mass market...right anexanhume ?
The rumors are that it went away because the market didn't make sense. Why make LCHBM when you can sell as much HBM2 as you can make?
 

Pheonix

Banned
Dec 14, 2018
5,990
St Kitts
People should stop undervolting the GPU and the CPU in these tests and instead decrease the boost clocks for the CPU/GPU. That way silicon lottery doesn't come into play. IMO those tests are not drawing the correct picture.

You can easily underclock a GPU by using MSI Afterburner by moving down the clock stages as a whole iirc
been saying the same thing myself. People shouldn't undervolt. Just downclock.
 

Deleted member 12635

User requested account closure
Banned
Oct 27, 2017
6,198
Germany
been saying the same thing myself. People shouldn't undervolt. Just downclock.
Yeah, just that kind of benchmarks of course.

For a daily system setup UV is a god's sent. Especially with AMD. I now have UV on my GPU at 975mV and -0.1V offset with my CPU. Saves a lot of power draw. My system is now at max 140W CPU and max 185W GPU for power draw. And that is achieved with a power hungry Vega 56.
 

sncvsrtoip

Banned
Apr 18, 2019
2,773
People should stop undervolting the GPU and the CPU in these tests and instead decrease the boost clocks for the CPU/GPU. That way silicon lottery doesn't come into play. IMO those tests are not drawing the correct picture.
You think that if he didn't undervolt the power consumption would be smaller ? Because 208w is still little too much for consoles. IIRC undervolting decrease power consumption.
 

Metalane

Member
Jun 30, 2019
777
Massachusetts, USA
FP16 precision is too low for the vast majority of graphics processing. Rounding errors accumulate fast and on a big TV screen become very apparent, less so on a smartphone sized displays which is why it's used widely in mobile gaming.
There are still many applications for FP16 though, if not in pixel shaders then in compute and simpler stuff like post processing. It will be used on next gen machines for sure but not to the extent of making them twice faster of course. Just some low percentage improvements on average here and there.
Thanks! Sounds great either way.
I am SOOO ready to join Team 20tf!!!... maybe in 10 years from now. ;)
We'll be in the PFLOPS range by then.
 

klik

Banned
Apr 4, 2018
873
So Navi 5700 is about 5% better than Vega 64/GTX 1080 in gaming performance on 4K according to Techpowerup review.


So we are already in 13TF GCN territory. Crazy that few months ago realistic expectation was 8-10TF GCN.


If year ago someone told me that we'll (PROBABLY) get 8 core Ryzen comparable to i7 9700k,13+TF GCN GPU,S faster than anything now SSD,Raytracing,3D Audio i would say they're dreaming.

Next gen is really gonna be glorious.
 

Deleted member 12635

User requested account closure
Banned
Oct 27, 2017
6,198
Germany
You think that if he didn't undervolt the power consumption would be smaller ? Because 208w is still little too much for consoles.
No, I think the power draw is maybe too optimistic because of the UV. But that is not my point. By UV you assume all chips could run at a same low voltage which is obviously not true. This is part of the silicon lottery. Trying to run your silicon on lower voltages just limits your yields.

To increase yields is the whole reasoning why the Hovis Method exists for the Xbox One X.
 

sncvsrtoip

Banned
Apr 18, 2019
2,773
No, I think the power draw is maybe too optimistic because of the UV. But that is not my point. By UV you assume all chips could run at a same low voltage which is obviously not true. This is part of the silicon lottery. Trying to run your silicon on lower voltages just limits your yields.
And ?;d He is not manufacturing navi cards, he try to simulate gonzalo apu so he undervolt to lower power consumption and keep console power limit.
 

AegonSnake

Banned
Oct 25, 2017
9,566
People should stop undervolting the GPU and the CPU in these tests and instead decrease the boost clocks for the CPU/GPU. That way silicon lottery doesn't come into play. IMO those tests are not drawing the correct picture.

You can easily underclock a GPU by using MSI Afterburner by moving down the clock stages as a whole iirc. Just click on the left symbol for the clock graph that will open a second window. There select the max clock speed point and press the "Shift" key while you moving the point up or down with the mouse. This way you can move the clock curve easily as a whole.
as if 8.95 tflops wasnt low enough already.

we are getting 7.9 tflops arent we.
 

Deleted member 12635

User requested account closure
Banned
Oct 27, 2017
6,198
Germany
as if 8.95 tflops wasnt low enough already.

we are getting 7.9 tflops arent we.
Not my intention to lower the computational performance of the consoles,I just think those benchmarks should be based on the correct assumptions.

I also think the next gen consoles will have at least 44CUs. That that prediction tier of mine is called "The Realist"
 
Last edited:

sncvsrtoip

Banned
Apr 18, 2019
2,773
And so what?

The purpose of those tests (I thought) was to get an impression how a 5700XT performs at lower clocks in regards of benchmarks and power consumption (at a similar setup like a Gonzalo is). So the variable is clock and not voltage.
He lower clock and voltage to better simulate gonzalo as it will be lower voltage on gonzalo than on 5700xt and 3700x. You would prefer only underclocking, I'm satisfied with his approach. Lets move on.
 

Metalane

Member
Jun 30, 2019
777
Massachusetts, USA
Well, prepare yourself because 10TF is at the very top end of what we can expect, and even then, only with N7+ or Samsung's EUV.
I just want 2080 performance. Some people here have said anywhere between 10-12 RDNA FLOPS would suffice that.
Data is meaningless if you don't know how to interpret it.
I never said how I determined it.
Just an observation. To see a sentence like " I can't believe we get anything below 10TF " tells me you are not entirely "data driven" and instead you want your expectations to be fulfilled ;)
The data statement was referring to my crazy Peta FLOPS prediction. As far as next gen TFLOPS go I honestly don't know. I don't think anyone in here knows tbh yet. We have all these numbers jumping back and forth. It's not a diss to this thread or anyone in here, it's just that we still don't have enough info to solidify a proper number.

Predictions: At least one platform holder will brag about how many gigarays their console is capable of. I'm guessing that a 36 CU RDNA2 GPU at 1800 MHz (8.29 TF) will weight in around 9 Gigarays per second, which would put it inbetween 2080 Super and 2080 Ti in terms of ray count.
So RTX level RT capabilities? This may sound weird but is there a good balance between general processing power and RT power in the RTX cards?

Edit: Rays are the new polygons?

Whats the point to release 6 AAA exclusives in one year? The there would be timespan where we dont get big exclusives at all and not to mention those games would eat each others sales and would need to complete against third party games too. It isn't happening, even If it was possible.

We will get 2-3 AAA games from Sony each year max.
Well Sony did say they wanted more AAA games didn't they?
 
Last edited:
Jul 6, 2018
174
Predictions: At least one platform holder will brag about how many gigarays their console is capable of. I'm guessing that a 36 CU RDNA2 GPU at 1800 MHz (8.29 TF) will weight in around 9 Gigarays per second, which would put it inbetween 2080 Super and 2080 Ti in terms of ray count.
 

gundamkyoukai

Member
Oct 25, 2017
21,148
Whats the point to release 6 AAA exclusives in one year? The there would be timespan where we dont get big exclusives at all and not to mention those games would eat each others sales and would need to complete against third party games too. It isn't happening, even If it was possible.

We will get 2-3 AAA games from Sony each year max.

If we look at Sony plan for PS4 first year it would have been around that number but a whole bunch of stuff got delay .
I expect both MS and Sony will try to have more than 2 or 3 AAA games the first year .
 
Jul 6, 2018
174
I just want 2080 performance. Some people here have said anywhere between 10-12 RDNA FLOPS would suffice that.

I never said how I determined it.

The data statement was referring to my crazy Peta FLOPS prediction. As far as next gen TFLOPS go I honestly don't know. I don't think anyone in here knows tbh yet. We have all these numbers jumping back and forth. It's not a diss to this thread or anyone in here, it's just that we still don't have enough info to solidify a proper number.


So RTX level RT capabilities? This may sound weird but is there a good balance between general processing power and RT power in the RTX cards?

Edit: Rays are the new polygons?


Well Sony did say they wanted more AAA games didn't they?

You can want 2080 raw performance all day long, but 2070/2060 Super is more likely. That said console optimizations will more than make up the difference in most AAA games.

I don't think nVidia went all in on ray-tracing with the RTX line-up. RT cores are estimated to add only 6-7% to the area of RTX GPUs. I think AMD's tech lends itself more naturally to ray-tracing support.
 

anexanhume

Member
Oct 25, 2017
12,913
Maryland
been saying the same thing myself. People shouldn't undervolt. Just downclock.
Both are appropriate.

1. Normal/Gaussian distributions apply. If most chips can hit 3.6GHz base and the higher binned SKUs are only 100 to 200 MHz higher on base clock, that tells you even the worst performing parts should hit 3.4GHz base at stock voltage. In reality, 3700X would need to cover the meat of the distribution and most parts can probably do 3.6GHz base. Undervolting is then appropriate if your max target clock is lower.
2. Yield will be driven by GPU performance. GPU will be taking up over 2/3 the die area and dissipation. It needs to have good performance to yield as a console part. The CPU will be on the same silicon quality, so it's performance will fall in line. The chances that a chip that can't hit 3.2GHz on an undervolted CPU can hit their GPU targets seems low.
 

AegonSnake

Banned
Oct 25, 2017
9,566
If we look at Sony plan for PS4 first year it would have been around that number but a whole bunch of stuff got delay .
I expect both MS and Sony will try to have more than 2 or 3 AAA games the first year .
Horizon, gt sports, Spiderman and maybe gow should all be released by late 2021 with maybe one of those games getting delayed to spring2022.

They better have a plan to release games from bluepoint, their new san Diego studio or some other third party studio like kojipro within the first year or two. Because just two games in all of 2021 won't be enough.
 

Conal

Banned
Oct 26, 2017
2,868
Sorry if this is an annoying question but do we have any examples of what Next Gen games are expected to look like at this point?
 
Status
Not open for further replies.