Question Speculation: RDNA2 + CDNA Architectures thread

Page 167 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

uzzi38

Platinum Member
Oct 16, 2019
2,702
6,405
146
All die sizes are within 5mm^2. The poster here has been right on some things in the past afaik, and to his credit was the first to saying 505mm^2 for Navi21, which other people have backed up. Even still though, take the following with a pich of salt.

Navi21 - 505mm^2

Navi22 - 340mm^2

Navi23 - 240mm^2

Source is the following post: https://www.ptt.cc/bbs/PC_Shopping/M.1588075782.A.C1E.html
 

Shivansps

Diamond Member
Sep 11, 2013
3,873
1,527
136
The 3090 is $800 more than the 3080 for roughly 10-15% more performance. If anything AMD's pricing is significantly better.

Also, the 6900XT (based on AMD's benchmarks) is more than 2x a 5700XT. A 5700XT scores 60fps at 4K max quality in Forza Horizon 4, a 6900XT scores 169fps in the same game at 4K.

Being cheaper than a card that is already waaaaaaaaaaaaaaaaaaaaaaaaaaaaaaay overpriced for the performance is not a lot. Any price looks good compared to the 3090.
 

zinfamous

No Lifer
Jul 12, 2006
110,806
29,556
146
Life pro tip: If your'e in the US and your credit is good the chase freedom flex card offers 200$ cash back ,if you spend 500$ in the first 3 months plus 5% cashback on Paypal purchases till the end of December plus 0% apr for 15 months.

That would make these cards 350$,420$, and 850$ respectively all for just a measly credit check.

yes, I have been considering that for IF the Pixel 4a 5g ever releases (my phone is beyond dead, and I don't want to splurge on better phones that won't work "As well" with Fi)

...now, hmm, yeah. not so sure, lol
 

gorobei

Diamond Member
Jan 7, 2007
3,713
1,067
136
Was 300w expected for 6900xt and 6800xt? Seems higher than my recollection of rumors

Pending reviews I will most likely end up getting a 6900xt. $999 is firmly in the range of acceptably overpriced for me
the rumors were that amd was waiting to see what ampere could do at what price. they then chose to push the clocks/power to try and beat nv on more areas, thus exceeding the original leaks on big navi power.
 

zinfamous

No Lifer
Jul 12, 2006
110,806
29,556
146
For the second time, read again the slides of testing methodology about RX 6800XT. Rage Mode is only for RX 6900XT.

Secondly. You accuse AMD of overplaying their GPUs, when AMD is showing absolute worst case scenario, since the AIB GPU will clock way higher than reference designs.

No I'm not, read what I said.

You have a...strange history of interpreting my honest observations as negative criticism, lol.

How many straight generations of AMD-only stuff, for the most part, have I owned now? lol

How much money did I make off of betting on AMD over the last 4 years?

Anyway^ that post was an early response to my first seeing that one slide, after lunch, without reading any of ~4 pages of posts before I made it, so chill out. I'm going back through the read and reading now, clarifying some things.

Anyway, SAM is still a big ??? into how that will work in the end. There is the argument to be made that, of course, overclocking or not, you still have to be in the AMD ecosystem to see these results (Zen 3, 550 MB...maybe 450+? we don't know yet), and how widespread that will be...again, as always "dependent on developer implementation."

DO NOT take my points as any kind of negative criticism to what is a very, very, very impressive release for AMD today. This is frankly better than I expected! (hoped a little cheaper...but it still beats nVidia)
 

zinfamous

No Lifer
Jul 12, 2006
110,806
29,556
146
I don't think it requires developers to program for it, I found that highly doubtful given they showed it's effects across 10 existing games that I highly doubt they patched for this feature. I think it's more that SAM's performance improvement will be highly variable from game to game and could be coded for to get max benefit. AMD didn't go into hardly any detail about the feature so we'll have to wait and see exactly how it works though.

right, because it is shown as a game effect here (but I am responding to a point in Ryan's article that I read about...2 minutes before making that comment):

Meanwhile, Smart Access Memory will have a degree of automatic support to it. As things stand, many AAA games can use it today automatically and get some degree of a performance gain. But to get the best results, developers will still want to program games with the technology in mind.

so yes, it works to some degree and is automatic in some titles now (I'm guessing this something that is far more likely with this gen because uh...well, I was going to say because of the fancy wizardry in the consoles, but those are all Zen 2, right? ...but they are custom, so the "automatic" SAM implementation, that we are seeing in these charts, is maybe something that exists in the custom Zen 2 + RDNA2 parts in PS5 and XBOXsSXS; but the "finer tuned" stuff that requires additional work, for mo better performance, is still that future grey area?).

However--what this means to me is that what they are showing is actually quite real NOW and what you can expect going forward...as a minimum (even though Rage mode performance, also in Ryan's article, seems a bit silicon lotteryish), and so now we have even further potential for performance beyond this if SAM really does get serious love, and as devs area already actually using it for real (instead of the historical point of everyone hoping they will, at some point, to you know, make GCN better, lol), this is very promising.

It's like we're back in that world where you can honestly expect to squeeze out another 10-15% performance from release throughout the life of these things with AMD driver updates and game optimizations.

So, I'm left wondering how much of that Rage + SAM boost is represented by each of those implementations?
 

ddogg

Golden Member
May 4, 2005
1,864
361
136
So glad to see the 6900XT being this competitive with the 3090. It's been I guess back in the AMD Thunderbird / 9700 Pro days (if you want to consider ATI as AMD for this purpose) since I last had an all AMD system as my main pc. I just hope I can manage to grab both a 6900XT and a 5950X on launch day. I'm needing all the computing muscle I can manage since getting this Samsung Odyssey G9 (which is as stunningly awesome as it's made out to be).

Pretty much the same as me. Last time I had an all red system was during the 9800 Pro days. I recently picked up a LG CX48 and my aging 1080Ti struggles at 4k.
I'm hoping I can get both the 5950X and 6900XT on launch day.
 
Reactions: Tlh97 and YBS1

zinfamous

No Lifer
Jul 12, 2006
110,806
29,556
146
Gamers Nexus reporting Rage Mode is not an overclock. It's just a mild power limit increase, which in turn allows for a slightly increased boost clock in games.

No lottery for this. I assume it's because they didn't want to release a card with >300W power limit.

Gamers Nexus also stating AMD has said Rage Mode only accounts for a 1-2% performance increase. AMD also claims smart memory access works automatically and the performance gains they have shown are simply from it being enabled, not because the games have support for it baked in.

ah OK, tp update my recent post, where I mentioned that Ryan also said this about Rage mode...I went back and read it, and see how I misread that. Here's the whole section, for context:

Automatic overclocking is not a new idea, even to AMD and its partners. So absent any further details on Rage Mode, I am curious how they’re implementing it and how they’re handling stability checks. Despite the one-click nature of automatic overclocking, I’ve found that previous implementations have had much higher friction in practice, especially when a system would crash during overclocking testing when it found an unstable GPU frequency. If AMD has a way to gracefully do this, that would greatly improve the experience, I believe.

Though regardless, users will need to keep in mind that any kind of one-click overclock stability testing will not be as rigorous as the stability testing done at the factory. So I don’t expect every last RX 6000 owner to use the feature, and I don’t expect it to put AMD’s partners out of business when it comes to factory overclocked video cards.

Meanwhile it should be noted that AMD has confirmed that Rage Mode is not warranty voiding. So users can still claim warranty coverage on their cards if they’ve used AMD’s official, user-friendly methods of overclocking.

The performance gains from overclocking, in turn, will vary with chip and board quality. Some chips barely make spec, others fly past it; so overclocking is always subject to the silicon lottery. To that end, AMD Isn’t offering any specific performance claims with respect to Rage Mode, but they are publishing a slide showing a combination of gains in their testing from Rage Mode and Smart Access Memory.

Basically, the first line of that fourth paragraph, I misread as still talking about Rage mode, because he used "Rage" and "Overclocking" a lot previously, and essentially called it overclocking in the line right before it: "AMD's official, user-friendly version of overclocking"

In this case, I see though, he's just talking about general overclocking when it comes to silicon lottery.

....I guess the demarcation is "voids warranty or doesn't" lol, if we want to not call Rage mode "overclocking." For now, Ryan is calling it that, but the other guys aren't. I guess the community will settle on something.



Still, what about SAM working on non 5xx boards, and of course it requiring Zen3 to even exist? (still catching up here, don't bite my head off, people )

I'm in the "no plan whatsoever to replace my 470X Taichi group" and depending on Q1 BIOS update, will strongly consider swapping my 2700X for a 5600X but otherwise, ...it's definitely less benefit for me...but considering that I'd be moving from a VEGA 64, I shouldn't complain anyway.
 
Last edited:

VRAMdemon

Diamond Member
Aug 16, 2012
6,572
7,823
136
It's really amazing how far AMD has come. Weren't they almost bankrupt like 4 years ago?. I have built many systems for family and friends going back to the 3dfx days and have never used anything AMD. Intel and Nvidia have always been the better buy performance wise. The performance boost when pairing a AMD cpu with an AMD gpu is really cool if pulled off. No wonder Nvidia wanted to get into cpu's.

I haven't built a system since the beginning of Z87 Haswell launch, but it's time to upgrade my system and the wife's. Our daughter is out of the house now and with no kids around, we will have more free time and plan to get back into gaming. So I plan on building 2 systems once the dust settles here and all this new hardware is readily available and somewhat maturing.

It's time to detach from the Intel teat. AMD is on a roll and the hardware looks good. Time to throw some money their way. One of my new builds will be AMD (maybe both) Looking forward to building my first AMD system! Pretty cool time to get back into PCs and gaming.
 

Glo.

Diamond Member
Apr 25, 2015
5,761
4,666
136
I have this bad feeling that if is $399 it is because is not that much faster than the 5700XT. Lets say it is 25% faster, at $399 that would put it way too close to the 6800, and it would likely match or be better than a 3070, NOTHING that AMD has done this year so far makes me belive they are going to give you more perf for the same price. They are more than capable of making the RX 6700XT $499 matching (or improving) 3070 perf while offering more ram and price the 6700 at $399.
As I have said in this thread: 40 CU GPU should be between RTX 2080 Super(10% above) and RTX 2080 Ti, 60 CU should be 10-20% above RTX 2080 Ti, 80 CU die should be 40-50% above RTX 2080 Ti.

So far the two last one were pretty spot on, so I don't think my info was wrong on the 40 CU die .
 

moinmoin

Diamond Member
Jun 1, 2017
4,994
7,765
136
I wonder how Raja feels now, or how much of this he expected.
I think he should have known of the general approach, like Zen 1 to show it is competitive with the competition, Zen 2 to show it can match it, Zen 3 it can surpass it. Now RDNA1 competitive, RDNA2 match, RDNA3 surpass? Maybe he didn't expect it to be executed so well, but he should have been aware that AMD under Dr. Su takes the best case scenarios of the competition as goal and works within that frame to set up its own targets like perf/W or IPC (which it then surpassed so far).
 

PhoBoChai

Member
Oct 10, 2017
119
389
106
I agree on the RX 6800, wa shoping $549 at most.
However, that tells me they are confident that it will cream the RTX 3070

It's already a cream.

Why would u want a 8GB GPU at 2080Ti perf levels for the next few years for $499, when you pay a tiny bit extra, you get +15% more performance AND 16GB VRAM?

The 3070 is the biggest loser here. No wonder NV rushing out another cut GA102 for 3070Ti.
 

PhoBoChai

Member
Oct 10, 2017
119
389
106
Ah lame. +$300 for a 6900XT doesn't seem remotely worth it then. They must be better binned chips to be hitting the same TBP, but I suspect enabling the same features between both and OCing to the gills will net a pretty minor dif between 6800/6900.

More likely, 6900XT is power limited, same bios 300W limit, more CUs. It's not going to boost much above "Game Clock". AIB cards with 350W limit or more, its gonna fly. On water for enthusiasts, especially.
 

Ajay

Lifer
Jan 8, 2001
16,094
8,106
136
It's already a cream.

Why would u want a 8GB GPU at 2080Ti perf levels for the next few years for $499, when you pay a tiny bit extra, you get +15% more performance AND 16GB VRAM?

The 3070 is the biggest loser here. No wonder NV rushing out another cut GA102 for 3070Ti.
Some of us have hard budgets to work with. Like, hey honey, is it okay if I spend around $500 on a new video card this winter? "... stay under $500 and we're good". Yes
New cards come out, wow, the performance is awesome! Wait, over $500, noooo
 

insertcarehere

Senior member
Jan 17, 2013
639
607
136
RDNA2 certainly looks to be a very competitive architecture from what AMD has shown, even with clocks that aren't stratospheric like some in this thread have insisted they would be, barely mentioning RT at all (performance or implementation) is pretty concerning, so hopefully there is more info when NDA is lifted.

Lineup wise, the $579 6800 being so cut down (60CU, 14 Tflops) makes it way closer to the XSX than it really ought to be, PC gaming feeling more and more like a sucker's game if this is the pricing we get from both OEMs going forward...
 
Reactions: Tlh97 and Ajay

Ajay

Lifer
Jan 8, 2001
16,094
8,106
136
Lineup wise, the $579 6800 being so cut down (60CU, 14 Tflops) makes it way closer to the XSX than it really ought to be, PC gaming feeling more and more like a sucker's game if this is the pricing we get from both OEMs going forward...
Yep. No game subsidies feeding GPU makers coffers. Heck, they often give away games to move product!
 
Reactions: spursindonesia
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |