WYP
News Guru
AMD's Multi-Chip RDNA 4 GPU Flagship appears to be a MONSTER - Too bad it was cancelled.
Read more about AMD's rumoured RDNA 4 flagship specifications.
Read more about AMD's rumoured RDNA 4 flagship specifications.
Beastly specs revealed for cancelled AMD RDNA 4 flagship
- Thread starter WYP
- Start date
NeverBackDown
AMD Enthusiast
Could have just been a prototype for testing and collecting data for such a complex design to prepare them for future more dense GPU chiplets.
AngryGoldfish
Old N Gold
Sounds like they made a wise move. I'd rather they got this thing right with RDNA5 unlike what they did with Navi 31.
Dicehunter
Resident Newb
RDNA4 will likely be around the same perf as the current RDNA3 lineup from what I've read but with much better power usage and at a vastly reduced price making really good PC gaming more in reach of many more people.
RDNA5 will likely be like what we saw with small Navi to big Navi i.e 5700XT to 6900XT. A really big jump in performance.
Not so much the price but mainly to do with power use and clocks. RDNA3 can match a 4090 in raster performance if you get a specially tweaked VBIOS, I think it was Der8auer who showed it last year. The downside is it puts the power draw up to 600+ watts which also then demands watercooling for the immense amount of heat.
AMD could've gone this route but they would've been a laughing stock.
RDNA5 will likely be like what we saw with small Navi to big Navi i.e 5700XT to 6900XT. A really big jump in performance.
Not so much the price but mainly to do with power use and clocks. RDNA3 can match a 4090 in raster performance if you get a specially tweaked VBIOS, I think it was Der8auer who showed it last year. The downside is it puts the power draw up to 600+ watts which also then demands watercooling for the immense amount of heat.
AMD could've gone this route but they would've been a laughing stock.
Last edited:
KingNosser
Well-known member
I think it's partly price as that is obvious but more importantly communication between all thr dies and software driver side where they need more time to get it right.
The 7900xtx could have been slightly better but a issue with the ark side of the die made it not fully fixable but still decent.
So my feeling is once amd feel they have solved all issues they are looking at and ready oh it'll work very well but I expect price is still a big factor, they have no desire to manufacture vast products if it's going to have issues they have learned alot lately.
Rdna4 will be better than a 7900xtx in general but at I expect slightly between 7800xt 7900xt pricing depending on current stock it took them a fair time to shift the 6k series stuff but they have I think it's learning not to be in the same situation.
I honestly feel with Intel slowly making progress and nvidia doing so well that amd simply want to force the market in the midrange solve the ark driver issues until it's ready maybe not even until rdna6 but when they are ready it will stomp.
The biggest issue they have is dxr but rdna4 will improve this it's the only area where they are behind. Thing is if you can get something viable at price performance and have a new 480era it's what amd need to take market share.
I'm pretty positive while the high end rdna4 products are cancelled the r&d side has continued and I feel it's so that at a later date you're see things combined better than current apus to heavily take the laptop market.
If they can make dies in and form they need and mix and match for purpose with drivers and ark working well they will be in a strong position.
No different to nvidia wanting a cpu of there own like arm.
Smaller dies with mix match less defects cheaper higher yields on even cheaper nodes are going to be key.
It's just taking time with getting it all to gel right I feel having so many dies needing to talk to each other or work out all the quirks can't be easy but once they do well then unless Intel and nvidia seriously do the same they will have issues.
Obviously nvidia has a huge software eco system in many ways this is there biggest advantage. But amd are selling a huge chunk of them new 300x cards so they aren't doing as bad as some might say.
Intel are I feel in huge trouble part reason I think they are investing in the newer node they need a winner and fast I feel but I doubt they will disappear or collapse they are a huge company but they need a winner and not had that for a while now.
Thou amd can't once in a good position revert or slack and get greedy otherwise they could end up at $2 a share again.
Intel thou massive issues nvidia thou can switch things up far easier and I expect more than ready to do so in terms of r&d already done years ago, but there better at making products for the time and profit than the absolute best they could make.
So a shift is happening but the company ladder I feel is as is for now.
The 7900xtx could have been slightly better but a issue with the ark side of the die made it not fully fixable but still decent.
So my feeling is once amd feel they have solved all issues they are looking at and ready oh it'll work very well but I expect price is still a big factor, they have no desire to manufacture vast products if it's going to have issues they have learned alot lately.
Rdna4 will be better than a 7900xtx in general but at I expect slightly between 7800xt 7900xt pricing depending on current stock it took them a fair time to shift the 6k series stuff but they have I think it's learning not to be in the same situation.
I honestly feel with Intel slowly making progress and nvidia doing so well that amd simply want to force the market in the midrange solve the ark driver issues until it's ready maybe not even until rdna6 but when they are ready it will stomp.
The biggest issue they have is dxr but rdna4 will improve this it's the only area where they are behind. Thing is if you can get something viable at price performance and have a new 480era it's what amd need to take market share.
I'm pretty positive while the high end rdna4 products are cancelled the r&d side has continued and I feel it's so that at a later date you're see things combined better than current apus to heavily take the laptop market.
If they can make dies in and form they need and mix and match for purpose with drivers and ark working well they will be in a strong position.
No different to nvidia wanting a cpu of there own like arm.
Smaller dies with mix match less defects cheaper higher yields on even cheaper nodes are going to be key.
It's just taking time with getting it all to gel right I feel having so many dies needing to talk to each other or work out all the quirks can't be easy but once they do well then unless Intel and nvidia seriously do the same they will have issues.
Obviously nvidia has a huge software eco system in many ways this is there biggest advantage. But amd are selling a huge chunk of them new 300x cards so they aren't doing as bad as some might say.
Intel are I feel in huge trouble part reason I think they are investing in the newer node they need a winner and fast I feel but I doubt they will disappear or collapse they are a huge company but they need a winner and not had that for a while now.
Thou amd can't once in a good position revert or slack and get greedy otherwise they could end up at $2 a share again.
Intel thou massive issues nvidia thou can switch things up far easier and I expect more than ready to do so in terms of r&d already done years ago, but there better at making products for the time and profit than the absolute best they could make.
So a shift is happening but the company ladder I feel is as is for now.
Dicehunter
Resident Newb
From everything I've read so far the 7900XTX will remain AMD's fastest card but RDNA4 will excel significantly in power use and price point at the low to mid range which is where most of the money is to be made.
We'll only see a true successor to the 7900XTX when RDNA5 lands which has me theorising that RDNA4 may just be a stop gap product.
KingNosser
Well-known member
Oh it's a stop gap alright sure high end on hold but to say it's fully cancelled isn't the truth just the current release of it.
When I say in general better I'm more talking the same as you we agree but where rdna4 has an advantage I expect due to amd and Sony working closely is in other tasks as the dxr side should be an improvement so performance wise in rasta similar but in dxr better by how much unsure.
Thou I also feel in fsr/frame gen also maybe improvements really depends on how much sharing sony and amd allow but it's known that they work closely which is why when Intel tried so hard with Xbox to get the contract that they missed out.
Just my general thoughts not like I'm saying the 8800xt will beat the 7900xtx by like 10% or anything like that just similar with benefits on ark.
The price thou I'm unsure what a 8800xt will be exactly but it's going to sell like hot cakes if it's under 600.
When I say in general better I'm more talking the same as you we agree but where rdna4 has an advantage I expect due to amd and Sony working closely is in other tasks as the dxr side should be an improvement so performance wise in rasta similar but in dxr better by how much unsure.
Thou I also feel in fsr/frame gen also maybe improvements really depends on how much sharing sony and amd allow but it's known that they work closely which is why when Intel tried so hard with Xbox to get the contract that they missed out.
Just my general thoughts not like I'm saying the 8800xt will beat the 7900xtx by like 10% or anything like that just similar with benefits on ark.
The price thou I'm unsure what a 8800xt will be exactly but it's going to sell like hot cakes if it's under 600.
Dicehunter
Resident Newb
No doubt it'll sell well if the price is right but from everything I've seen so far it won't match the 7900XTX in performance but rather its strength will be vastly reduced power use and the performance for the price will be excellent due to it being aimed at the low to mid range.
I'm looking at RDNA4 as a possible sneak peek of what RDNA5 will be.
I'm looking at RDNA4 as a possible sneak peek of what RDNA5 will be.
KingNosser
Well-known member
That's the thing next gens mid range is close to the current high end but as said its in dxr ect other tasks where I expect it to perform better so in my view not far off at all a 7900xtx.
Bit like the 6800xt vs 7800xtx the 7800xt performs better but in async tasks the 6800xt is ahead it'll be similar really depends just how many CUs the 8800xt will have.
It's going to be a popular card since least on amds side nothing above it unless they switch the general naming around but by solely focusing on low mid far less risk and waste.
Like we all know thou the price is the key factor.
Nvidia with the 5090/80 of course are going to be far ahead but I feel until PS6 ect is out there are few games going to need such a card. Plus if your able to play everything would such a card as a 5090/80 this year even make sense when by the time it's needed 6000 series be out at similar prices and far stronger.
Like all times in tech always a sweet spot generally 2nd gen of new nodes as pro amd as ive been these last few years to me it doesn't matter if it's Intel or amd or even a rebirth of voodoo it wouldn't matter who but bringing a card to market that brutally forces the change in price is much needed and that is me saying even if it was nvidia that decided to do so the prices simply must normalise now.
The 5090 is going to be extremely expensive and sure as a card for the task best but in all seriousness a total waste. I'd even find it hard to justify a 4090 right now unless 4k as there aren't that many games that push it hard at all very few.
So a card at the right price that plays everything is all that's needed outside of enthusiasm towards it it's a hobby after all.
I even expect the 5070 to out perform the 8800xt that amd will bring but I do hope that it's either worth my interest or that amd are going to get me excited again for the 9000 series.
Cause atm I think the ps5 and 6800xt pc I have is fine and it's most likely going to be zen6 where I should seriously think on a refurbishment.
Still it's only maybe 6 months until we're know for certain and I'd rather smd work out the kinks of multiple dies fully than have another vega moment they been doing well last few gens and something like that would be a costly mistake.
Bit like the 6800xt vs 7800xtx the 7800xt performs better but in async tasks the 6800xt is ahead it'll be similar really depends just how many CUs the 8800xt will have.
It's going to be a popular card since least on amds side nothing above it unless they switch the general naming around but by solely focusing on low mid far less risk and waste.
Like we all know thou the price is the key factor.
Nvidia with the 5090/80 of course are going to be far ahead but I feel until PS6 ect is out there are few games going to need such a card. Plus if your able to play everything would such a card as a 5090/80 this year even make sense when by the time it's needed 6000 series be out at similar prices and far stronger.
Like all times in tech always a sweet spot generally 2nd gen of new nodes as pro amd as ive been these last few years to me it doesn't matter if it's Intel or amd or even a rebirth of voodoo it wouldn't matter who but bringing a card to market that brutally forces the change in price is much needed and that is me saying even if it was nvidia that decided to do so the prices simply must normalise now.
The 5090 is going to be extremely expensive and sure as a card for the task best but in all seriousness a total waste. I'd even find it hard to justify a 4090 right now unless 4k as there aren't that many games that push it hard at all very few.
So a card at the right price that plays everything is all that's needed outside of enthusiasm towards it it's a hobby after all.
I even expect the 5070 to out perform the 8800xt that amd will bring but I do hope that it's either worth my interest or that amd are going to get me excited again for the 9000 series.
Cause atm I think the ps5 and 6800xt pc I have is fine and it's most likely going to be zen6 where I should seriously think on a refurbishment.
Still it's only maybe 6 months until we're know for certain and I'd rather smd work out the kinks of multiple dies fully than have another vega moment they been doing well last few gens and something like that would be a costly mistake.
