xool
Member
Was this debunked?
mmh
I just noticed that the MP2888A uses "NVIDIA's Open VReg specification" which is an open standard, but I couldn't find any evidence of AMD ever having used it ..
Needs checking
Was this debunked?
Consoles chips are a semi custom design they can be on any node they choose to be.when i think about i dont think 7nm+ is gonna be a thing next gen. the cpu is zen 2 which is regualar 7nm and 7nm + is for zen 3. How are gonna have a cpu with 7nm and gpu that is 7nm+ and make them compatabile???
Next-gen consoles will most likely use a monolithic APU design (judging by Scarlett), which means you can only use one process for the entire chip.when i think about i dont think 7nm+ is gonna be a thing next gen. the cpu is zen 2 which is regualar 7nm and 7nm + is for zen 3. How are gonna have a cpu with 7nm and gpu that is 7nm+ and make them compatabile???
(1.6GHz / 3.2GHz)Uhm interesting, maybe gonzalo it's a laptop apu ?
The AtariBox ?It would be funny if it was something random like an Alienware steam machine.
It would be really funny if Gonzalo it's not next gen related ahahaha XD All this talk for nothing!
of course it's not related there's to much discrepancies to suggest that it is and still some people spin other way around.It would be really funny if Gonzalo it's not next gen related ahahaha XD All this talk for nothing!
of course it's not related there's to much discrepancies to suggest that it is and still some people spin other way around.
i've never believed in gonzalo or second sku for xbox it's fan fiction, from rumor leaks.
No, its almost certainly related.of course it's not related there's to much discrepancies to suggest that it is and still some people spin other way around.
i've never believed in gonzalo or second sku for xbox it's fan fiction, from rumor leaks.
Zen2 was OPN in January, so 6 months. But be aware that devs got final APU dev kits in January last time around, so for consoles I would think its different and this is most likely a norm.Yes something is not right, OPN means qualification sample so it should be soon launched...it doesn't make any sense :/
Yes something is not right, OPN means qualification sample so it should be soon launched...it doesn't make any sense :/
of course it's not related there's to much discrepancies to suggest that it is and still some people spin other way around.
i've never believed in gonzalo or second sku for xbox it's fan fiction, from rumor leaks.
True but remember it's up to Sony & Ms not AMD .
GameCache™ (L3 Cache) has been cut down from the 32MB in Matisse. Looks like there'll be 16MB total for both CCXes.
* Chip looks to be using 16x8Gbit RAM chips (16GB total).
* Memory latency is very bad at 154ns. About 2x what you'd get on Zen 2 with decent DDR4. This is expected if you use GDDR6 as it trades latency for bandwidth.
I disagree (not Gonzalo) - also lower performance because clocks I think... which is almost certainly the same chip as Gonzalo
[..]
* CPU performance is a good chunk below even Zen 1.0 in those tests. About ~75-80% of a 1700X.
https://pastebin.com/y8qXme7b This was posted ? :/
Erm. Doesn't Matisse have 16MB per 4 core CCX too (with no connection between L3 caches) - ie no change
Yep looks slower than DDR4 (so GDDR obv.). Not sure how the benches give exact memory layout though?
I disagree (not Gonzalo) - also lower performance because clocks I think.
I also commented that this looks like a pre-Zen2 part based on float vs int performance - which contracticts Cerny's statement on PS5 being Zen2 - it's hard to tell but I think the multicore float performance shows it is inferior to Zen2 and more like Zen+
Yeah you're right - I was reading it wrong 2700X results match 8MB per CCX L3, 3700 results match 16MB per CCX L3, and Flute seems to correspond to 4MB L3 available to a core.Compare Flute
![]()
versus 3700X (2x16MB):
![]()
and you can clearly see the L3 cache is lower. 3700X can't access the L3 of the other CCX other than through the IO die, hence the big spike after 16MB.
Here's a 2700X (2x8MB):
![]()
The increase at 8MB vs 2700X is interesting though, I think it's probably more like 4MB per 2 cores. So 4x4MB.
Sorry saw that after I posted - yeah it's right there in black and white.Unknown U... 16GB
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 MHz
1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024 MB
Is a pretty dead giveaway that there's 16x8Gbit modules. Now I very much doubt consoles will have 512-bit buses so we can deduce that it's likely a 256-bit bus with 2 chips connected to each memory controller.
Erm. Doesn't Matisse have 16MB per 4 core CCX too (with no connection between L3 caches) - ie no change
Yep looks slower than DDR4 (so GDDR obv.).Not sure how the benches give exact memory layout though?[edit - seen it - that's 16 1024MB chips - that's a 512 bit bus - and a lot of bandwidth]
I disagree (not Gonzalo) - also lower performance because clocks I think.
I also commented that this looks like a pre-Zen2 part based on float vs int performance - which contracticts Cerny's statement on PS5 being Zen2 - it's hard to tell but I think the multicore float performance shows it is inferior to Zen2 and more like Zen+
Next gen Open World games should look like this. Or not? This match original watch dogs reveal? Or is Better?Been playing Detroit recently, and it looks amazing considering it's running on a base PS4. Next gen is gonna be insane!
A 14.2TF beast in 2020 will just about be able to hold its own against a 40TF PC in 2028.
Uhm interesting, maybe gonzalo it's a laptop apu ?
Seriously, that would be like bringing a water pistol to a tank fight.
Sony's fortunes are heavily dependant on PlayStation.
The new consoles must offer compelling software until 2028 and beyond.
The best way to do that is with high performance hardware from the initial release date. The majority of the user base will be gaming on the base model. Games are taking longer to develop due to increased complexity and high fidelity asset creation. The console lifecycle will also extend, not shrink, to offer a stable target platform that these huge titles can rely upon to be static from concept, through to release and beyond.
A 14.2TF beast in 2020 will just about be able to hold its own against a 40TF PC in 2028. A 8-9TF weakling, not so much.
If you still think we're getting weak sauce 9TF/16GB/256 bit bus consoles then I don't know what to tell you.
Like bringing 1.2TF machine vs 4TF in 2013. I would look at hard limits of console design (TDP, die size) before going onto "omg they CANNOT have less then 15TF I dont care how they do it".
Because if they went with 9TF that would be top end of what AMD can provide. Much better then back in 2013 when you could have MUCH better GPU then 1.2-18TF ones
Sony's fortunes are heavily dependant on PlayStation.
The new consoles must offer compelling software until 2028 and beyond.
The best way to do that is with high performance hardware from the initial release date. The majority of the user base will be gaming on the base model. Games are taking longer to develop due to increased complexity and high fidelity asset creation. The console lifecycle will also extend, not shrink, to offer a stable target platform that these huge titles can rely upon to be static from concept, through to release and beyond.
A 14.2TF beast in 2020 will just about be able to hold its own against a 40TF PC in 2028. A 8-9TF weakling, not so much.
If you still think we're getting weak sauce 9TF/16GB/256 bit bus consoles then I don't know what to tell you.
There will not be a 14.2TF console next year. At least not in RDNA Flops.
But, it seems like 1.8GHZ is sweet spot for RDNA and going down to 1.4.-1.5 wont result in expected TDP gain as in years of GCN.On this I feel we must agree to disagree.
Slow, wide, power efficient, 7nm EUV, 14.2 Navi TF.
Sony's fortunes are heavily dependant on PlayStation.
The new consoles must offer compelling software until 2028 and beyond.
The best way to do that is with high performance hardware from the initial release date. The majority of the user base will be gaming on the base model. Games are taking longer to develop due to increased complexity and high fidelity asset creation. The console lifecycle will also extend, not shrink, to offer a stable target platform that these huge titles can rely upon to be static from concept, through to release and beyond.
A 14.2TF beast in 2020 will just about be able to hold its own against a 40TF PC in 2028. A 8-9TF weakling, not so much.
If you still think we're getting weak sauce 9TF/16GB/256 bit bus consoles then I don't know what to tell you.
We learn from history that we don't learn from history.
This isn't like PS4 where Sony's back was against the wall after the huge misstep with the PS3.
PS5 has more in common with the PS2 launch. Huge, profitable customer base on the back of the previous console's success, backwards compatibility to lock in the users to the ecosystem. What did we get with the PS2? A great performance jump at a fantastic price.
isn't navi 10 lite should mean cut down version of 5700xt or 5700? but you suggest it's 40 cu's like xt version wich doesn't fit "lite' meaning.BTW now 3 leaks (Gonzalo, PCB and this new one) tie in perfectly, independently.
As in, Gonzalo and userbenchmark leak show same APU, that is Navi 10 Lite and 3.2GHZ 8 core CPU. While UB leak gives us CPU score, it also gives us memory (clamshell) arrangement which means 16GB of GDDR6 (look at latency yikes) on 256bit bus. You can see 33GB/s speed for chip, which means memory speed is >16Gbps (and only Samsung 18Gbps bring that, 16GB/s is max 32GB/s).
This in turn ties in with PCB leak that specified 32GB of RAM in dev kit with Samsung 18Gbps chips on 256bit bus, which would result in 512GB/s - 528GB/s.
I am now 99% certain Gonzalo/PCB is legit and, since UB leak tells us L3 cache is halved, full 40CU + RT and 8 core Zen2 on 256bit bus would fit in 316mm² die (again, number that would tie in with exact number from PCB leak - leaked before Zen2 and Navi where shown and before we knew die sizes for each).
As suggested by Apisak (back in January) Navi 10 could have been aimed as big chip (40CU+) but was scrapped as mid sized one. 10 for AMD is used for first chip on market, doesnt have to be biggest, but usually is. So plan may have been changed to putting out mid tier GPU while Navi 10/Lite remained.isn't navi 10 lite should mean cut down version of 5700rx or 5700? but you suggest it's 40 cu's like rx version wich doesn't fit "lite' meaning.
How did previous generations came out ? ... exactly, i think you should learn from history how big companies operate .
If we pursue your logic, why Microsoft launched a weaker console ??????????????
knowing that they have more money than Sony ...
Ms's back was not against the wall, so it means Ms wanted to shoot herself on the foot, also you cant compare Ps2 with todays market, we are not talking about the same parts .
40Tf PC's in 2028 ... what tha hell are you smoking .
BTW now 3 leaks (Gonzalo, PCB and this new one) tie in perfectly, independently.
As in, Gonzalo and userbenchmark leak show same APU, that is Navi 10 Lite and 3.2GHZ 8 core CPU. While UB leak gives us CPU score, it also gives us memory (clamshell) arrangement which means 16GB of GDDR6 (look at latency yikes) on 256bit bus. You can see 33GB/s speed for chip, which means memory speed is >16Gbps (and only Samsung 18Gbps bring that, 16GB/s is max 32GB/s).
This in turn ties in with PCB leak that specified 32GB of RAM in dev kit with Samsung 18Gbps chips on 256bit bus, which would result in 512GB/s - 528GB/s.
I am now 99% certain Gonzalo/PCB is legit and, since UB leak tells us L3 cache is halved, full 40CU + RT and 8 core Zen2 on 256bit bus would fit in 316mm² die (again, number that would tie in with exact number from PCB leak - leaked before Zen2 and Navi where shown and before we knew die sizes for each).
Id say 8.3TF - 9TF. 9TF would be absolute max I would expect of these consoles.How many tf would this amount to?
Uhm interesting, maybe gonzalo it's a laptop apu ?
...Or devkits with actual silicon will have to be provided relatively soon (beginning of 2020 certainly) as it was always the case. You cannot have QS chip for console half a year before launching console like you would with ordinary CPU/GPU.Well its not Sony or Microsoft. This is a qualification sample, so its got to be for a product soon launching.
...Or devkits with actual silicon will have to be provided relatively soon (beginning of 2020 certainly) as it was always the case. You cannot have QS chip for console half a year before launching console like you would with ordinary CPU/GPU.
No, dev kits with final chips. OPN is generally found in chips 6-7 months before retail (in GPU/CPU space) but you would certainly want to have final APU 7-8 months before retail in console space because, unlike retail GPUs and CPUs, you have to provide it to actual developers.Final devkits 1,5 years before launch? That would be a first.