AMD Radeon RX 5950 XT - Finally a 2080 Ti Killer

erek

Supreme [H]ardness
Joined
Dec 19, 2005
Messages
4,410
Will it be enough to topple the impending RTX 3080 Ti series from the Green Team?

"Optimized 7nm+ process node
  • Enthusiast-grade desktop graphics card options
  • Hardware-Level Ray Tracing Support
  • A mix of GDDR6 and HBM2 graphics cards
  • More power-efficient than First-Gen Navi GPUs
Nothing else is stated, but from the looks of it, this could be a very powerful graphics card, marking AMD's proper return in the enthusiast high-end desktop GPU space. We also know for a fact that AMD is planning to introduce hardware-accelerated ray tracing as a primary featured of their next-generation RDNA powered GPUs. With that said, the information is based on a rumor so treat it with a grain of salt, but if it is true, then we might get to hear at least some form of confirmation from AMD at CES 2020 aside from the expected Radeon RX 5600 XT announcement."

https://wccftech.com/amd-radeon-rx-navi-21-gpu-2x-performance-5700-xt-die-size-rumor/
 
  • Like
Reactions: viivo
like this

Thunderdolt

Limp Gawd
Joined
Oct 23, 2018
Messages
427
Why would AMD confirm a 2021 launch of a new platform at a tradeshow held less than 7 days into 2020? And do so before their competition has even announced their 2020 launch?

Honestly, I'm not sure why this is even posted on [H]. wccf seems to be using a bot to autonomously write blurbs for clicks and there should be a defacto ban on reposting their garbage. What's newsworthy about a "rumor" that AMD's next generation will be improved over their current generation? You know, I've got some super secret inside information that says Zen2's successor is going to be even better than Zen2. It's going to use an improved process node, it's going to have various different clock speeds, AMD will sell it with a few different core counts, and it will utilize one or more of the two available types of RAM. Now here are a couple of links to similar ruminations on nothing elsewhere on my site in order to manage SEO and keep suckers clicking.
 
Joined
Apr 11, 2017
Messages
524
Why would AMD confirm a 2021 launch of a new platform at a tradeshow held less than 7 days into 2020? And do so before their competition has even announced their 2020 launch?

Honestly, I'm not sure why this is even posted on [H]. wccf seems to be using a bot to autonomously write blurbs for clicks and there should be a defacto ban on reposting their garbage. What's newsworthy about a "rumor" that AMD's next generation will be improved over their current generation? You know, I've got some super secret inside information that says Zen2's successor is going to be even better than Zen2. It's going to use an improved process node, it's going to have various different clock speeds, AMD will sell it with a few different core counts, and it will utilize one or more of the two available types of RAM. Now here are a couple of links to similar ruminations on nothing elsewhere on my site in order to manage SEO and keep suckers clicking.
Well if we waited for launch for everything it would be pretty dead. I love reading tech news everyday, but not everyone launches products or puts out reviews everyday either, talking about upcoming products is a great way to fill in those gaps. Although I agree Wccf is a cesspool. I check their site everyday, but I wish their comment section was more like here.
Feel free to post your own legit news stories...
 

erek

Supreme [H]ardness
Joined
Dec 19, 2005
Messages
4,410
Well if we waited for launch for everything it would be pretty dead. I love reading tech news everyday, but not everyone launches products or puts out reviews everyday either, talking about upcoming products is a great way to fill in those gaps. Although I agree Wccf is a cesspool. I check their site everyday, but I wish their comment section was more like here.
Feel free to post your own legit news stories...
Why would AMD confirm a 2021 launch of a new platform at a tradeshow held less than 7 days into 2020? And do so before their competition has even announced their 2020 launch?

Honestly, I'm not sure why this is even posted on [H]. wccf seems to be using a bot to autonomously write blurbs for clicks and there should be a defacto ban on reposting their garbage. What's newsworthy about a "rumor" that AMD's next generation will be improved over their current generation? You know, I've got some super secret inside information that says Zen2's successor is going to be even better than Zen2. It's going to use an improved process node, it's going to have various different clock speeds, AMD will sell it with a few different core counts, and it will utilize one or more of the two available types of RAM. Now here are a couple of links to similar ruminations on nothing elsewhere on my site in order to manage SEO and keep suckers clicking.
 

Snowdog

[H]F Junkie
Joined
Apr 22, 2006
Messages
10,067
I like how they can't even do basic math. How do they double Navi 10 die size and only get 15-16 Billion Transistors. It would obviously be ~20 Billion.

Erek, could you avoid posting low quality sources like WCCFT?

If you want a good quality source of leaks, and industry info, it's Videocardz.com.

There will be a lot less stories, but they will be of higher quality and veracity.
 

erek

Supreme [H]ardness
Joined
Dec 19, 2005
Messages
4,410
I like how they can't even do basic math. How do they double Navi 10 die size and only get 15-16 Billion Transistors. It would obviously be ~20 Billion.

Erek, could you avoid posting low quality sources like WCCFT?

If you want a good quality source of leaks, and industry info, it's Videocardz.com.

There will be a lot less stories, but they will be of higher quality and veracity.
i did post Buildzoid's YouTube response to this debacle above also
 

limitedaccess

Supreme [H]ardness
Joined
May 10, 2010
Messages
7,567
Why would AMD confirm a 2021 launch of a new platform at a tradeshow held less than 7 days into 2020? And do so before their competition has even announced their 2020 launch?
I'm not confirming the validity of this rumor but what you are pointing out by itself does preclude their being a basis to it due to past precedent.

AMD at CES 2018 confirmed that 7nm Vega would be sampling in 2018Q4. This was eventually launched at CES 2019 as Radeon VII.

Also not sure why you mention a 2021 launch, the rumor article does mention it would launch later in 2020.

Polaris was revealed at CES 2016, launched half a year later. Vega was revealed at CES 2017, launched 8 months later.

Over the last few years AMD has used CES to promote products launching very far out.
 
  • Like
Reactions: N4CR
like this

Jandor

Gawd
Joined
Dec 30, 2018
Messages
527
Will it be enough to topple the impending RTX 3080 Ti series from the Green Team?

"Optimized 7nm+ process node
  • Enthusiast-grade desktop graphics card options
  • Hardware-Level Ray Tracing Support
  • A mix of GDDR6 and HBM2 graphics cards
  • More power-efficient than First-Gen Navi GPUs
Nothing else is stated, but from the looks of it, this could be a very powerful graphics card, marking AMD's proper return in the enthusiast high-end desktop GPU space. We also know for a fact that AMD is planning to introduce hardware-accelerated ray tracing as a primary featured of their next-generation RDNA powered GPUs. With that said, the information is based on a rumor so treat it with a grain of salt, but if it is true, then we might get to hear at least some form of confirmation from AMD at CES 2020 aside from the expected Radeon RX 5600 XT announcement."

https://wccftech.com/amd-radeon-rx-navi-21-gpu-2x-performance-5700-xt-die-size-rumor/
This is the thing reported on chinese networks in November by guys working compatible hardware who visited TSMC facilities. So they saw tests of Navi 21 and have
knowledge of its size. Doesn't mean it was ready. In fact it was kind of not ready then with 300W+ TDP. Furthermore they seem to say it's on GDDR6 for what they have seen.
So this is not fresh news at all !

Also this is 7nm not 7nm+ which is bad. 7nm is far from 100% yield at this size, more close to 50%. If they don't go HBM2 is on the contrary of what wcctech says, not because of the package and the size of the GPU but because GDDR6 is still way cheaper now and they want their GPU to be cheaper to produce than the 2080Ti.
 

SmokeRngs

[H]ard|DCer of the Month - April 2008
Joined
Aug 9, 2001
Messages
16,060
I like how they can't even do basic math. How do they double Navi 10 die size and only get 15-16 Billion Transistors. It would obviously be ~20 Billion.

Erek, could you avoid posting low quality sources like WCCFT?

If you want a good quality source of leaks, and industry info, it's Videocardz.com.

There will be a lot less stories, but they will be of higher quality and veracity.
Stop attacking the poster and stop reading threads and clicking on links you don't like instead of telling someone not to post something. He's not hiding the source and he's not trying to trick anyone.
 

ChadD

Supreme [H]ardness
Joined
Feb 8, 2016
Messages
4,533
Why would AMD confirm a 2021 launch of a new platform at a tradeshow held less than 7 days into 2020? And do so before their competition has even announced their 2020 launch?

Honestly, I'm not sure why this is even posted on [H]. wccf seems to be using a bot to autonomously write blurbs for clicks and there should be a defacto ban on reposting their garbage. What's newsworthy about a "rumor" that AMD's next generation will be improved over their current generation? You know, I've got some super secret inside information that says Zen2's successor is going to be even better than Zen2. It's going to use an improved process node, it's going to have various different clock speeds, AMD will sell it with a few different core counts, and it will utilize one or more of the two available types of RAM. Now here are a couple of links to similar ruminations on nothing elsewhere on my site in order to manage SEO and keep suckers clicking.
This source aside.

AMD has talked about rDNA2... its slated for Q4 2020 not 2021. All they have confirmed for 100% sure... is the target launch being late 2020, hardware based RT, and that they are going after the high end. Based on Lisa Su comments in their last earnings call rDNA2 sounds like its a 7nm+ part. They haven't confirmed they are chiplet and they may not be... but I sort of lean toward they are as rDNA 2 will be in the SOC for the next gen consoles that feature chiplet based zen 2. Which would make me disbelieve any rumors of it being a massive die.
 

Imhotep

Gawd
Joined
Feb 12, 2014
Messages
812
I don't think its such a crazy idea that it will be a big die. Consoles will get a cutdown version of this and chiplet designs are beyond RDNA2
 

RanceJustice

Supreme [H]ardness
Joined
Jun 9, 2003
Messages
6,091
This source aside.

AMD has talked about rDNA2... its slated for Q4 2020 not 2021. All they have confirmed for 100% sure... is the target launch being late 2020, hardware based RT, and that they are going after the high end. Based on Lisa Su comments in their last earnings call rDNA2 sounds like its a 7nm+ part. They haven't confirmed they are chiplet and they may not be... but I sort of lean toward they are as rDNA 2 will be in the SOC for the next gen consoles that feature chiplet based zen 2. Which would make me disbelieve any rumors of it being a massive die.
I'm looking forward to it. Its kind of a pity they're scheduled for the later part of 2020, given how long it has been since 5700XT's release , but if they can really do it right then all will be forgiven. Hardware RT (preferably with an open spec , similar to those chips used in Xbox Scarlett and PS5 will help to cement an open RT application vs Nvidia's typical proprietary preferences), targeting the high end of what will be out at the time and forthcoming, and ideally reasonably priced - will definitely be a winner. If it destroys the NV 3080 Ti that's great, but if they have models that compete favorably with it while costing significantly less they'd have a winner ; same goes for a step-down card that goes up against the 3080 and 3070 etc.
 

SmokeRngs

[H]ard|DCer of the Month - April 2008
Joined
Aug 9, 2001
Messages
16,060
I'm looking forward to it. Its kind of a pity they're scheduled for the later part of 2020, given how long it has been since 5700XT's release , but if they can really do it right then all will be forgiven. Hardware RT (preferably with an open spec , similar to those chips used in Xbox Scarlett and PS5 will help to cement an open RT application vs Nvidia's typical proprietary preferences), targeting the high end of what will be out at the time and forthcoming, and ideally reasonably priced - will definitely be a winner. If it destroys the NV 3080 Ti that's great, but if they have models that compete favorably with it while costing significantly less they'd have a winner ; same goes for a step-down card that goes up against the 3080 and 3070 etc.
I'm not surprised that it's going to be a while for anything higher than the 5700XT. As it is that card drinks too much power and puts out too much heat. It was going to require a process refinement and/or a respin of the silicon to get the architecture to a point where it will scale higher at reasonable power draws. I think this is the exact reason the 5700XT was the first released card and we haven't seen anything higher performance. AMD knew that was going to be the best it could do with the first iteration of the architecture and left it there instead of trying to put out a card which was going to suck down way too much power and put out too much heat for a price which was too high.
 

ManofGod

[H]F Junkie
Joined
Oct 4, 2007
Messages
11,531
Without a doubt, the high end will be considerably faster than the 5700XT by the time it is released. However, I am content with what I already have but others will not, that is the cool thing about competition and being able to upgrade whenever you want too.
 

Snowdog

[H]F Junkie
Joined
Apr 22, 2006
Messages
10,067
I'm looking forward to it. Its kind of a pity they're scheduled for the later part of 2020, given how long it has been since 5700XT's release , but if they can really do it right then all will be forgiven. Hardware RT (preferably with an open spec , similar to those chips used in Xbox Scarlett and PS5 will help to cement an open RT application vs Nvidia's typical proprietary preferences), targeting the high end of what will be out at the time and forthcoming, and ideally reasonably priced - will definitely be a winner. If it destroys the NV 3080 Ti that's great, but if they have models that compete favorably with it while costing significantly less they'd have a winner ; same goes for a step-down card that goes up against the 3080 and 3070 etc.
This is not about open or closed API for RT HW.

RT is done with Microsoft DXR API. The GPU makers support the API.
 

defaultluser

[H]F Junkie
Joined
Jan 14, 2006
Messages
12,994
Complete horseshit.

You can't magically add double the number of flops plus high-performance dedicated RT plus the extra die space consumed by HBM2 bus interface, and pretend like it will be anywhere near DOUBLE the die size pf Navi 10 !

That baby would be almost 700mm ^2.


There's a reason nobody mixes both HBM and GDDR6 on the same die: they are completely different connection technologies, which means a lot of unused elements wasting die space. This is the reason why every iteration of Nvidia's Big Volta features HBM2, and AMD's Vega derivatives all use HBM2.

To believe this Wonder-drug of a video card can even exist is silly. Cards can mix GDDR5 and GDDR6 without much added complexity, because the bus width are the same, it's just the control logic that's different.
 
Last edited:

Lakados

[H]ard|Gawd
Joined
Feb 3, 2014
Messages
1,957
I heard the new AMD GPU runs at sub zero temperatures natively and actively cools your PC. While managing to offer 10 times the performance while giving you lap dances. Sadly happy endings are a subscription service not available in Quebec and Utah.
 

Mode13

Gawd
Joined
Jun 11, 2018
Messages
750
I heard the new AMD GPU runs at sub zero temperatures natively and actively cools your PC. While managing to offer 10 times the performance while giving you lap dances. Sadly happy endings are a subscription service not available in Quebec and Utah.
Wow it'll be available in Nebraska?

 

Lakados

[H]ard|Gawd
Joined
Feb 3, 2014
Messages
1,957
Wow it'll be available in Nebraska?

Yeah AMD contacted the necessary senators and heads of state in advance and gave them advice on what VPN’s to use to connect to the service to keep their wives and political rivals in the dark on them using it. Which they get for free of course.
 

limitedaccess

Supreme [H]ardness
Joined
May 10, 2010
Messages
7,567
Again as a disclaimer this has nothing to do with the validity of this specific rumor.

In terms of launch history -

Pitcairn/Tahiti (2012, small/mid, GCN) -> Hawaii (2013, large, GCN update) - > Tonga (2014, mid, GCN update) -> Fiji (2015, large, GCN update) -> Polaris (2016, mid, GCN update) -> Vega (2017, large, GCN update) -> Navi (2019, mid, RDNA) -> ? (2020, large, RDNA update)

It certainly wouldn't be surprising if they launched a large GPU using an RDNA iteration in later 2020.

Pitcairn - 212mm^2, 2.8B Transistors. "190W." 20CU. 32 ROP. 256 bit. 2.56 TFLOPs.

Hawaii - 438mm^2, 6.2B Transistors. "300W." 44CU. 64 ROP. 512 bit. 5.6 TFLOPs

Just as an aside Hawaii had some additional functionality over Pitcairn as well such as 1/2 FP64. I wouldn't assume how much RT hardware support takes up (transistors/die budget) or that it will be of similar scope to Nvidia's. Turing's die size increase isn't just due to RT hardware support (but this is a bit of complex side topic)
 

defaultluser

[H]F Junkie
Joined
Jan 14, 2006
Messages
12,994
Again as a disclaimer this has nothing to do with the validity of this specific rumor.

In terms of launch history -

Pitcairn/Tahiti (2012, small/mid, GCN) -> Hawaii (2013, large, GCN update) - > Tonga (2014, mid, GCN update) -> Fiji (2015, large, GCN update) -> Polaris (2016, mid, GCN update) -> Vega (2017, large, GCN update) -> Navi (2019, mid, RDNA) -> ? (2020, large, RDNA update)

It certainly wouldn't be surprising if they launched a large GPU using an RDNA iteration in later 2020.

Pitcairn - 212mm^2, 2.8B Transistors. "190W." 20CU. 32 ROP. 256 bit. 2.56 TFLOPs.

Hawaii - 438mm^2, 6.2B Transistors. "300W." 44CU. 64 ROP. 512 bit. 5.6 TFLOPs

Just as an aside Hawaii had some additional functionality over Pitcairn as well such as 1/2 FP64. I wouldn't assume how much RT hardware support takes up (transistors/die budget) or that it will be of similar scope to Nvidia's. Turing's die size increase isn't just due to RT hardware support (but this is a bit of complex side topic)

No, Hawaii only jumped up yo 1/8 FP64 (MUCH smaller hit on die size thann1:2 your rambling memory ). Pitcarin already had 1:16, so it wasn't much of a jump; the HD 7970 remains the FP64 beast of it's generation!

The reality is that doubling your fp units AND your memory bus takes more than double the transistors. I've already gone into expected die size increase from the dedicated RT units AND the jump up to 4096-bit bus lines.

There is added overhead in adding more compute units to a chip, so the scaling size is always "a little more than double." This is pretty consistent with Pitcarin to Hawaii.
 
Last edited:

limitedaccess

Supreme [H]ardness
Joined
May 10, 2010
Messages
7,567
No, Hawaii only jumped up yo 1/8 FP64 (MUCH smaller hit on die size thann1:2 your rambling memory ). Pitcarin already had 1:16, so it wasn't much of a jump; the HD 7970 remains the FP64 beast of it's generation!

The reality is that doubling your fp units AND your memory bus takes more than double the transistors. I've already gone into expected die size increase from the dedicated RT units AND the jump up to 4096-bit bus lines.

There is added overhead in adding more compte units to a chip, so th e scaling size is always "a little more than double"
Hawaii itself was capable of 1/2 FP64, like many consumer SKUs it was segmented down to 1/8 for R9 290/x. Pro variants were fully enabled for 1/2 FP64. Pitcairn was only capable of 1/16, including pro implementation.

I made no comment regarding scaling but Hawaii itself was more than double the size (slightly) but more than double the CUs (although double the memory controllers, ROPs). As another comparison point GP104 314mm^2 -> GP102 471mm^2 is exactly 50% more. A sizable amount of GPU die space is also other hardware that doesn't need to scale up.

As for RT support and the additional die space, I don't want to get into it too much but I feel most people have an over estimation of how much die space it actually adds due to misconceptions with respect to Turing. Also as I said RT support is a loose term as opposed to absolute, AMD's implementation could (and I suspect likely) will be much lower in scope than Nvidia's. Non-RTX Turing (eg. 1660ti) for instance has stronger relative RT support than Pascal, it performs closer to the GTX 1080 in RTX but is normally more comparable to the GTX 1070.

But I'll state again I'm not making a specific comment with regards to scaling or the plausibility of these speculations. It's just that they should be simply dismissed as implausible at least based on precedent and other technical aspects.
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,272
Why would AMD confirm a 2021 launch of a new platform at a tradeshow held less than 7 days into 2020? And do so before their competition has even announced their 2020 launch?

Honestly, I'm not sure why this is even posted on [H]. wccf seems to be using a bot to autonomously write blurbs for clicks and there should be a defacto ban on reposting their garbage. What's newsworthy about a "rumor" that AMD's next generation will be improved over their current generation? You know, I've got some super secret inside information that says Zen2's successor is going to be even better than Zen2. It's going to use an improved process node, it's going to have various different clock speeds, AMD will sell it with a few different core counts, and it will utilize one or more of the two available types of RAM. Now here are a couple of links to similar ruminations on nothing elsewhere on my site in order to manage SEO and keep suckers clicking.
Who supports banning even fake news? Only a tyrant suggests such asinine ideas. Let people judge for themselves what is good for them to read. I like wccftech, sometimes most of the time ugh thier bad. But I want options, not some control freak banning everything they dont like for me to see.
 

SmokeRngs

[H]ard|DCer of the Month - April 2008
Joined
Aug 9, 2001
Messages
16,060
WTFTech is bs "rumors" aka shit people made up. its not reliable and never has been.
There is a report button under his posts. If his posts are so offensive to you then click that button and report the post. One of two things will happen, the posts will be removed and he will likely be told why or you will be removed and your threadcrapping will stop.

No one has made you read the post. No one has forced you to post in the thread. His posts are little different from [H] News posts when the site was up. He hasn't broken a single rule. Here is the link to the rules for posting threads in this subforum since you don't know what they are and need to be shown he hasn't broken any rules.
 

pendragon1

[H]F Junkie
Joined
Oct 7, 2000
Messages
15,997
There is a report button under his posts. If his posts are so offensive to you then click that button and report the post. One of two things will happen, the posts will be removed and he will likely be told why or you will be removed and your threadcrapping will stop.

No one has made you read the post. No one has forced you to post in the thread. His posts are little different from [H] News posts when the site was up. He hasn't broken a single rule. Here is the link to the rules for posting threads in this subforum since you don't know what they are and need to be shown he hasn't broken any rules.
where do i say im offended? why did you read my post? why did you feel the need for this rant? wouldnt the same apply to you and my post?! why are you white-knighting? see the report button below my post? hit it if you want.
edit: when [H] was up they didnt use WTFTech as a source for "news" and i never said he broke rule, just posted garbage.
 

Mega6

2[H]4U
Joined
Aug 13, 2017
Messages
2,745
didnt think it was, you didnt quote me. was i not supposed to respond either? if anything, posting crap copy/pasted from garbage rumor sites will kill it.
I've already stopped posting news previously anyways - and won't be posting new articles anymore. Too many baby whiners here.
 

DeathFromBelow

Supreme [H]ardness
Joined
Jul 15, 2005
Messages
7,316
posting crap copy/pasted from garbage rumor sites will kill it.
Is there an article or actual documentation of their dishonesty/failures in the past? I'm honestly curious about their backstory. This isn't the first time I've heard these sort of things said about them.

I don't mind wild rumors as long as they're disclosed as such. Not like there's a whole lot of interesting tech to talk about lately.
 

Thunderdolt

Limp Gawd
Joined
Oct 23, 2018
Messages
427
Who supports banning even fake news? Only a tyrant suggests such asinine ideas. Let people judge for themselves what is good for them to read. I like wccftech, sometimes most of the time ugh thier bad. But I want options, not some control freak banning everything they dont like for me to see.
I didn't suggest banning fake news. Is this one of those posts where you're going to come back to it in the morning and wonder just how high you were when you made it?
 

blackmomba

Limp Gawd
Joined
Dec 5, 2018
Messages
149
Is there an article or actual documentation of their dishonesty/failures in the past? I'm honestly curious about their backstory. This isn't the first time I've heard these sort of things said about them.

I don't mind wild rumors as long as they're disclosed as such. Not like there's a whole lot of interesting tech to talk about lately.
Don't know the exact story but they were caught changing some "prediction" they had made to make it seem that they were right or something like 3 years ago. Since then it's been cool to say they aren't reliable
 
Top