Iemand nieuwsgierig naar wat Patrick Schur allemaal geschreven heeft op het Duitse 3DCenter Forum? Dat dacht ik al:

Tekst door Patrick Schur is cursief
https://www.forum-3dcente...=12464185&postcount=14213
@dargo Why 5 watts too much? 255 watt TGP is right.
The AIBs are currently testing different VBIOS. There are also cards with a TGP of 280 W. But the 255 W seems to be the standard.
> OgrEGT How high is the clock at 280W?
I honestly don't know. The AIBs always talk about TGP and I've always assumed they mean everything by that. That's why I never asked. After all, does it make the most sense, doesn't it?
Unfortunately, I can't tell you how high the clock is at 280 W.
>rogame
> Zitat:
> Navi21 XL depending on the model (OC & non OC)
> Base clock 1800MHz to 1950MHz
> Boost clock around 2100MHz to 2200MHz (artificially limited like RX 5700? )
> I'm not sure if these are final
> Zitat:
> Navi21 XT
> Base clock 2000MHz to 2100MHz
> Boost clock around 2400MHz
My statement has thus been confirmed. But it seems to have the clock rates of another card here.
After a short consultation, I can now say that AMD means the power loss of the GPU and GDDR6 with the TGP. Everything else that is on the map is not included there.
I am not bound by any NDA. Of course I can talk about the number of pieces if I knew them.
@Kriegsgeier This leaker is reading here by the way.
Incidentally, I didn't confuse Navi 21 with Navi 22.
The custom card has a TGP of 255 W. But I can't tell you whether it was overclocked and the consumption is actually higher or the ~ 2.4 GHz was achieved with 255 W TGP. Hope this is a little more understandable.
Navi 23 is not Apple exclusive and will come for laptops.
Didn't Apple want to rely entirely on ARM in the future? At least I know nothing about Apple exclusive hardware. If you don't work for Apple or AMD directly, it will be pretty hard to find out about it.
I don't know anything about the number of CUs yet. As Igor said, it is really hard to get information about Big Navi right now. AMD tries to prevent a premature leak by all means. For example, the AIB partners recently had a driver that "hid" the clock and limited performance.
@Piefkee What do you mean? I already said that the drivers limited the functionality of the card. I can't tell you exactly what AMD did either. This should probably prevent benchmarks from being leaked with real performance.
Perhaps I should have defined "recently" more precisely ... The partners have had the right driver since around the beginning of September. Of course, the driver is still being worked on continuously.
@horn 12 You don't have to mark my name in red. I've already read your post and I don't know when the launch will be. I don't think AMD will miss the Christmas business.
Does anyone have a PCGH account and can I tell them that I don't mean total consumption with TGP? I really don't know where they got it from. I can't think of anything more about PCGH ...
Edit: Hopefully it's done now.
> Originally from HOT
> So high consumption and all previous "performance data" are for the bin.
What previous performance data do you mean? I've only ever talked about the TGP and it even agrees with Igor's numbers. But here it also calculates the TBP. Whether his information on the TBP is correct or not is an open question.
I'll quote myself.
Source: https://twitter.com/patrickschur_/st...55621157392393
Quote:
- Max. TGP of Navi 21 XT is 290 W.
- AIB partners are testing different TGP for their cards.
- The ~ 2.4 GHz was achieved with an AIB partner card. So assume that the card was slightly overclocked.
By that I mean that AIB partners are currently testing cards with a TGP of 255W - 290W.
> Quote:
> Igor wrote that the driver is severely restricted.
I already said that yesterday.
> Quote:
> Is that still the case?
No, since the beginning of September they have a beta "unlock" driver. Of course, the driver is still being tuned vigorously.
Oh, it's nice to read along here ... All that's missing is a pack of popcorn. ;D
> Quote:
> By the way, the BIOS comes from Patrick Schur.
https://www.igorslab.de/d...bios-verraet-die-details/
> Originally from HOT
> Will be the Strix I would say. Or something completely sick from MSI.
If I remember correctly, I sent him a VBIOS from an XFX card. After all, I didn't come up with the ~ 2.4 GHz.
> @Patrick
> What is then in N22 as max. GPU clock in the PP table? 3Ghz? Do you by any chance already have a
> bios ready?
Unfortunately I don't have a vBIOS yet.
As far as I can tell, the 255 W TGP is the standard on the XT. As already mentioned, 290 W TGP are possible with OC. To be very precise, it's 289 W (for the Krentenkakkers among you)
. According to AMD, more is not allowed. It also seems that with all XT cards, the max. Clock with 2577 MHz is specified in the vBIOS. With the XL the max. Clock specified with 2475 MHz. Standard TGP seems to be 203 W and a maximum of 230 W.
So when I look at the TGP, it's between standard and max. Allowed TGP always ~ 13.3%.
I already stated in one of my tweets that the AIBs are experimenting with different TGPs. For the XL this is 203, 220 and 230 W.
WTF? If I already say that the game clock on a card is ~ 2.4 GHz. How do you get to 2.8 GHz? The 2.8 GHz are probably the maximum that you can set in Wattman. But certainly not a single card will achieve that. It's like the 3000 MHz limit in Ryzen Master for the Infinity Fabric. Has nothing to do with reality. The GFX maximum for all XTs is 2577 MHz (according to vBIOS).
Can we all calm down now and come back to the actual topic? Is it always like this here? I thought we are in the Navi Thread and I only read something from Nvidia RTX the whole time
> Originally Posted by Linmoum
> Then make it change. ;D
If someone makes me an admin, that's fine. 
I read out the 290 W (aka 289 W) vBIOS for you. As already mentioned, the max. Clock also here at 2577 MHz.
https://imgur.com/50STzO6
All XT cards have these high clock rates stored in vBIOS.
> patrickschur
> I read out the 290 W (aka 289 W) vBIOS for you.
> Cyberfries:
> So now we know what to do if we want new information:
> Simply spam the navigation thread with nVidia.
I'm sorry, but I thought it would be better to have one picture than several. I can't tell you whether the fan settings are already read correctly.
@Cyberfries Please not ... unless you want to get rid of me
> Originally Posted by dargo
> I mean, what can you expect from the performance even with 4-5% more clock speed?
There should actually be a difference of ~ 4% between the 255 W and 290 W model (measured in a benchmark). 
@dargo I have no clock rates yet. With the exception of the ~ 2.4 GHz game clock (also measured in a benchmark).
I would be interested in your opinion. What do you think of the current rumors and what do you think the final specifications will look like?
In any case, I am very curious to see how the clock rates will turn out in the end. The rumor mill is currently assuming significantly low clock rates. Even if I can't quite understand it. In any case, we can only hope that AMD did not cause the AIBs to make a mistake and manipulated the vBIOS.
I think that's very unlikely. After all, the AIBs have to test their cards and the driver was already activated. The vBIOS (if not manipulated) at least confirms my assumption that we will be dealing with a GPU that is designed for very high clock rates.
> Der Korken: lang verhaal over infinity cache, verschillende clock speeds, allemaal onzin, ik extrapoleer toch wel met TFLOPS, etc.
@Der_Korken Thank you for the detailed answer.
> Nazar: The fact is and remains: you cannot sell what you do not have (if you are not on the stock exchange)!
@Nazar Of course I can sell something that I don't have yet ...
> Originally Posted by Elite_Warrior
> Patrick himself thinks the board is plausible (gaat over de coreteks leak van de die shot, op het igorslab forum)
https://coreteks.tech/art...ads/2020/10/featured2.jpg
The hole spacing is also correct and it is very similar to other evaluation boards. The die is of course complete cheese. He said that in one of his videos. This is just a montage.
> Nazar: I spoke of not having ... you of not having yet ... a small but important difference.
@Nazar Touché.
The ~ 2.4 GHz were certainly not permanently on. Hence the approximate sign in front of the number.
The clock rate is likely to have moved somewhere between 2.3 - 2.4 GHz (pure speculation). But isn't that the point of the game clock? It indicates the frequency that is most common during a game or benchmark.
> nairune: It would be interesting to see what kind of benchmark that was where the 2400 MHz were. If it
> was something like TimeSpy, the 255W BIOS wouldn't be bad.
@nairune It wasn't Timespy.
> Originally from Digidi
> 2577 is a damn grumpy number could this be a code to reveal leakers?
Since each XT vBIOS has a max. Clock of 2577 MHz, that can't be.
> Originally Posted by Dural
> There are still no performance drivers ... Only useful leaks come from China anyway.
I can't think of anything else ... Then where are all the China Leaks? The performance drivers are also already available. Apparently someone is not listening to me.
> Adam D.: Can someone explain to me why Patrick compares the 3070-N21XT? To exactly counter the rumor,
> Big Navi can only compete with the 3070? In the meantime, it was quite foreseeable that AMD had more to
> offer.
Why did I refer to the 3070? Quite simply because the VideoCardz article was about the 3070 and not the 3080.
> Originally Posted by Denniss
> And Igor and Helm from More Power Toll already had three other VBios so what?
If he had already had a vBIOS there, he wouldn't need mine either. Please switch on your head beforehand.
> Quote:
> ASUS ROG RX 6800 XT GAMING OC may be clocked at
> 1527/2160/2410
If the numbers are correct, we are already at almost 2200 MHz game clock and up to the 2400 MHz mark it is only ~ 9%. So what sounds unrealistic about that? If I may also remind you of the RX 5700 XT, some cards have also boosted up to their boost clock. With OC, there was even more to it.
@mironicus Here the same again from me
https://pbs.twimg.com/med...dD?format=jpg&name=medium
> patrick: So what sounds unrealistic about that?
> Cyberfries: Alright, everything's fine. We are happy about everything you share.
> Whether 40mhz or not ... that is not even 2% in the order of magnitude anyway.
@Cyberfries Thanks. I don't know what clock rate he was expecting, but there are also cards with a 2 GHz boost.
> Are you suggesting that N21 would have reached 10000pt with a throttled driver?
The 10,000 points were achieved with the unthrottled beta driver.
Edit: Oh, you refer to the other tweet. If that's true, you can count on even more points.
> Linmoun (in reactie op die shot door patrick) But with you one sees less of everything. You can measure
> roughly and tell us whether > = 500mm² is cold or warm.
There is no die on the original.
> Quote from Cyberfries
> Why is your picture sharper? Coreteks shows a bit more around the chip, where the other two have a
> black / gray square.
> The big question is really what is hiding in the middle.
> You probably don't want to show a heat spreader with the name of the AIB, understandable and
> meaningful.
> Whether Coreteks scaled the center of the picture - in other words: is the die fitted in its
> original size?
> Do not answer! - I want to be able to sleep tonight.
I will hide it from you.
spoiler
My image is sharper because I have the original. Coreteks shows nothing about the die. He inserted the GDDR6 himself and the die does not correspond to reality. He also said in his video.
What is hidden in the middle is a small circuit board with contacts. The carrier board with the die is placed on the contacts. The whole thing is an EV board by the way.ok ok TL:DR;
- 255 en 2,4 GHz (game clock) misschien toch niet gelinked.
- XT kaarten tussen de 255-290W. Max. clock altijd 2577 MHz.
- XL kaarten tussen de 203-230W. Max clock altijd 2475 MHz.
- geen XTX TGP of clocks.
- theoretisch max clocks van 2,8 GHz kunnen niet gehaald worden, dat is net als 3000 MHz in Ryzen Master voor de infinity fabric. niet realistisch.
- 289W misschien wel gekoppeld aan 2577MHz.
https://imgur.com/50STzO6
- "The rumor mill is currently assuming significantly low clock rates. Even if I can't quite understand it." ik niet hoor, Patrick

- "If the numbers are correct, we are already at almost 2200 MHz game clock and up to the 2400 MHz mark it is only ~ 9%. So what sounds unrealistic about that?" precies
- "My image is sharper because I have the original. Coreteks shows nothing about the die. He inserted the GDDR6 himself and the die does not correspond to reality. He also said in his video."
"What is hidden in the middle is a small circuit board with contacts. The carrier board with the die is placed on the contacts. The whole thing is an EV board by the way."
En hij is "as we speak" actief:
https://www.forum-3dcente...h.php?do=finduser&u=39010
[
Voor 8% gewijzigd door
sunsmountain op 23-10-2020 00:26
]