What is Radeon RX 7000 series for you? Please elaborate

[Dr. Dro]

It failed to meet every expectation. It's not an entirely bad product and I almost got one with the lowered prices, but the drivers are almost always in terrible condition and Radeon's always lacked the cool features of GeForce. I've reached a point in my life where I'm no longer comfortable in buying something and waiting (number of months/years) for me to enjoy it.

 

[GerKNG]

a heavily rushed product with lots of hardware bugs, software issues and by far not enough performance/efficiency.
we laughed about NVidia pulling 100W more than RDNA2 for similar performance. now it's the opposite.

i'll wait for RDNA4

 

[Macro Device]

drivers are almost always in terrible condition

Same could be said about the green ones. Especially from the no-control-panel DCH drivers era. Drivers stability is on par with nVidia nowadays, the issue is terrible marketing and almost complete lack of perf/w improvements.

buying something and waiting (number of months/years) for me to enjoy it

My 6700 XT is something I had to be wating for it to become enjoyable but it's not an AMD's fault. It's AIB's cooler. After modding, it's perfectly fine. Never had any issue in any game and software except for those caused by my unstable CPU which has nothing to do with AMD and nVidia anyway.

 

[Dr. Dro]

Same could be said about the green ones. Especially from the no-control-panel DCH drivers era. Drivers stability is on par with nVidia nowadays, the issue is terrible marketing and almost complete lack of perf/w improvements.

Could have fooled me...

I really wish AMD fans stopped playing this game of pretend that everything is fine and that Nvidia's not a billion light years ahead of them, it's getting really old. As much as we may have a problem with Ada's pricing, it's... not even a contest anymore. RDNA 3 barely shows gains, if any, against its own predecessor, while Nvidia literally figured out on the fly shader execution reordering to match the architecture's most favorable condition. That's how far the gap between them has grown, and that's why Nvidia's product stack is so suboptimal, they can release whatever really.

If anything I'm gonna go a step beyond... the 7900 GRE showing regressions v. Navi 21 on a similarly sized configuration (matching compute units, bus width, etc.) is something that should never have happened - yet, according to the benchmark figures that have been floating around, it's precisely what happened. The 6900 XT beats it and in some cases so does the 6800 XT. Same deal with Navi 33 v. Navi 23... yet this time verified even here on TPU, the 7600 is like, an utter disappointment, unable to do 5% over its fully enabled previous counterpart, the 6650 XT. If this keeps up, AMD might as well hang up their hat and go into retirement. Soon Intel won't be just an upstart, is AMD really willing to become the third option here?

How can one even say that the drivers are fine if release after release comes out completely botched and broken only on AMD graphics? Take Ratchet and Clank, it had tons of severe issues on AMD and that needed drivers and cooperation between the devs and AMD to get it barely functional, and ray tracing is still not working - again on AMD graphics alone. Man, I don't even care anymore.

 

[oxrufiioxo]

I really wish AMD fans stopped playing this game of pretend that everything is fine and that Nvidia's not a billion light years ahead of them, it's getting really old. As much as we may have a problem with Ada's pricing, it's... not even a contest anymore. RDNA 3 barely shows gains, if any, against its own predecessor, while Nvidia literally figured out on the fly shader execution reordering to match the architecture's most favorable condition. That's how far the gap between them has grown, and that's why Nvidia's product stack is so suboptimal, they can release whatever really.

I think the reason the Nvidia stack is so odd is the specifications where decided likely 2 years ago during a crypto boom.

I think billion light years is a bit of an exaggeration though haven't done any RDNA3 based builds because nobody I do builds for is interested in them but I did do plenty of 6800XT/6700XT/6600XT based builds and drivers have been fine. At worst they required a fully clean install of windows when switching from an Nvidia card and most games perform fine some better some worse the typical spread. I have recommended against multi monitor but a couple of them run dual screens without issues.


Like I said in a previous post this is the biggest disappointment to me






Not even close.

 

[Dr. Dro]

I think the reason the Nvidia stack is so odd is the specifications where decided likely 2 years ago during a crypto boom.

I think billion light years is a bit of an exaggeration though haven't done any RDNA3 based builds because nobody I do builds for is interested in them but I did do plenty of 6800XT/6700XT/6600XT based builds and drivers have been fine. At worst they required a fully clean install of windows when switching from an Nvidia card and most games perform fine some better some worse the typical spread. I have recommended against multi monitor but a couple of them run dual screens without issues.


Like I said in a previous post this is the biggest disappointment to me


View attachment 306971

View attachment 306972

Not even close.

I get what you mean, but to defend my point. the one thing I never denied about Ada was its architectural finesse, compared to what even Ampere used to be. In that regard... when you compare RDNA 3 which brings next to nothing new to to the table vs. its predecessor yet couldn't meet its own conservative marketing targets to Ada, that's what I mean by a billion light years ahead. Hey, at least the encoder doesn't suck so much anymore, no thanks to the community incessantly asking AMD for like 5 years to fix it... The performance gap may not be there (4090 notwithstanding) but only because the 40 series product stack is so bad.

 

[oxrufiioxo]

I get what you mean, but to defend my point. the one thing I never denied about Ada was its architectural finesse, compared to what even Ampere used to be. In that regard... when you compare RDNA 3 which brings next to nothing new to to the table vs. its predecessor yet couldn't meet its own conservative marketing targets to Ada, that's what I mean by a billion light years ahead. Hey, at least the encoder doesn't suck so much anymore, no thanks to the community incessantly asking AMD for like 5 years to fix it... The performance gap may not be there (4090 notwithstanding) but only because the 40 series product stack is so bad.

I'm still optimistic.... AMD first generation chiplet cpu had issues and have improved by leaps and bounds every generation. Maybe I am being naïve but I am hoping for a similar progression. I've said this 100 times but we need all 3 gpu vendors to be competitive or else progress will stagnate. Just like on the CPU side its best when both AMD and Intel offer compelling products at every price point.

 

[Vayra86]

In my experience they are a success. I did hover over so-so for a while, but realized I did so only because of one point: the initial MSRP. It was too high. But thats not saying anything about the hardware.

Now though, if you can snatch a 7900XT at around 800, or an XTX at 950 tops you are buying the best high end price/perf you can get at this time. On top of that, its solid stuff all the way through. There is some mild OC potential here. The out of the box game experience is just 100%. Newer drivers are touch ups to newer games at best.

Youre not waiting for whatever FSR or DLSS nonsense either because even a 7900XT just slaughters most things anyway. Temps and noise are in check. It is what I need and expect a gpu to do.

 

[ARF]

AMD needs the performance crown (or being close to it every generation) that can only be approached when they have a large chip >700 mm^2.
Why can nvidia make so large chips, while AMD can't?
TU102 was 754 mm^2.
AMD's Navi 10, Navi 31 and Vega 20 are 250-335 mm^2 chips.

 

[GhostRyder]

I was in the 'So-So' category, for context I would put the RX 6000 series as a success and the last few series before the 6000 as a disappointment. For me, the problem this generation comes to a few major points that hold them back.

1: The 7900 XTX being the top end only really competing and besting the RTX 4080.
2: The pricing structure for the 7900 XTX is not terrible compared to the RTX 4080, however the 7900 XT is such a huge drop off and at the beginning was not enough of a price reduction to warrant the huge loss in performance
3: The naming structure sucks (Can we please just have a standard naming structure, XT means higher and nothing means lower, why do we have to have XTX as meaning higher than XT)
4: Slow roll out of the mid range and lower end cards.

I am a fan of AMD and right now my card of choice would probably be a 7900 XT at the moment due to current pricing, I wish they would work harder next gen with making themselves competitive at the very high end like the 6000 series. I don't see software as a problem or lack of features as a problem as a lot of that stuff is limited in its uses anyways (Talking about game support for these features). How many times in the past we have been told that new features are going to be the best thing ever and will be widely adopted by everyone just for it to either be dropped or something else replace it shortly after. That is why when I normally look at video cards, I go based off the basic performance more than anything.

 

[AnotherReader]

AMD needs the performance crown (or being close to it every generation) that can only be approached when they have a large chip >700 mm^2.
Why can nvidia make so large chips, while AMD can't?
TU102 was 754 mm^2.
AMD's Navi 10, Navi 31 and Vega 20 are 250-335 mm^2 chips.

It isn't as if they can't make large chips. Fiji, used in Fury X and Fury, was 596 mm^2 and the MI200 is 724 mm^2. I suspect they don't think they will be able to sell large dies in meaningful volumes.

 

[oxrufiioxo]

It isn't as if they can't make large chips. Fiji, used in Fury X and Fury, was 596 mm^2 and the MI200 is 724 mm^2. I suspect they don't think they will be able to sell large dies in meaningful volumes.

I don't think they believe they can sell a 12-1500 gpu is probably why they decided on a 300mm2 GCD.. Only the most die hard AMD fans would likely buy it.

 

[AnotherReader]

I don't think they believe they can sell a 12-1500 gpu is probably why they decided on a 300mm2 GCD.. Only the most die hard AMD fans would likely buy it.

Exactly. A large die would have to sell for high prices, and they know that the people willing to shell out that much are more likely to buy Nvidia.

 

[oxrufiioxo]

Exactly. A large die would have to sell for high prices, and they know that the people willing to shell out that much are more likely to buy Nvidia.

Yeah, I mean look at how much the 7900 series had to drop they are almost competing with a tier lower gpu in pricing at this point. I guess if they could have released something 10-15% faster than a 4090 in raster they could have priced it around the 4080.... Obviously that wasn't appealing to them. I think they should have just to say they have the raster performance crown honestly assuming power consumption wasn't a meme.

 

[AnotherReader]

Yeah, I mean look at how much the 7900 series had to drop they are almost competing with a tier lower gpu in pricing at this point. I guess if they could have released something 10-15% faster than a 4090 in raster they could have priced it around the 4080.... Obviously that wasn't appealing to them. I think they should have just to say they have the raster performance crown honestly assuming power consumption wasn't a meme.

Given how much power a 4090 can consume in ray tracing, I don't think power consumption is a problem for the high-end market. AMD could have matched or beaten the 4090 with a 120 CU RDNA3 SKU, but it would still be behind in ray tracing.

 

[Dr. Dro]

I'm still optimistic.... AMD first generation chiplet cpu had issues and have improved by leaps and bounds every generation. Maybe I am being naïve but I am hoping for a similar progression. I've said this 100 times but we need all 3 gpu vendors to be competitive or else progress will stagnate. Just like on the CPU side its best when both AMD and Intel offer compelling products at every price point.

I was optimistic until I saw monolithic Navi 33 suffering from the same problems and inefficiencies... Either way they have their work cut out for them. And that means we as customers have to put the pressure on them instead of treating this multi billion dollar corporation as our darling.

Given how much power a 4090 can consume in ray tracing, I don't think power consumption is a problem for the high-end market. AMD could have matched or beaten the 4090 with a 120 CU RDNA3 SKU, but it would still be behind in ray tracing.

They couldn't, it doesn't scale that far. Besides they couldn't beat AD103 with N31 XTX as it was.

 

[gffermari]

AMD needs the performance crown (or being close to it every generation) that can only be approached when they have a large chip >700 mm^2.
Why can nvidia make so large chips, while AMD can't?
TU102 was 754 mm^2.
AMD's Navi 10, Navi 31 and Vega 20 are 250-335 mm^2 chips.

The 7900XT and XTX dies are 529sq.mm while the 4080 is 379sq.mm.
And the 4090 is 609...
How much bigger could AMD have made it?

 

[AnotherReader]

I was optimistic until I saw monolithic Navi 33 suffering from the same problems and inefficiencies... Either way they have their work cut out for them. And that means we as customers have to put the pressure on them instead of treating this multi billion dollar corporation as our darling.



They couldn't, it doesn't scale that far. Besides they couldn't beat AD103 with N31 XTX as it was.

It scales perfectly fine. The 7900 XTX scales pretty well from the 7900 XT. The off chips interface consumes a substantial amount of power so a 120 CU XTX would not have consumed 25% more power. The problem is that despite consuming over 400 W, this hypothetical SKU would have been far behind in ray tracing and so would have to be sold at a price significantly less than the 4090.

 

[Deleted member 229121]

The biggest thing AMD did to piss me off this round was happily subscribe to Jensenomics.
All they did was look at the performance #s and just slot them into the BS framework laid down by Nvidia.

Being the value gaming underdog was always their draw (with an occasional surprise left hook).
Even if their 2x offerings are cheaper, they're not exactly great deals either if you take a step back and look at the big picture.

I still have a 1440p G-Sync only monitor.

Same. I was all annoyed about it earlier on in the year, but honestly whatever now.
I'm going to play at 1440p till 2025 at this rate when something more affordable can handle 4K.

 

[oxrufiioxo]

I was optimistic until I saw monolithic Navi 33 suffering from the same problems and inefficiencies... Either way they have their work cut out for them. And that means we as customers have to put the pressure on them instead of treating this multi billion dollar corporation as our darling.



They couldn't, it doesn't scale that far. Besides they couldn't beat AD103 with N31 XTX as it was.

That's mostly due to the 7600 being stuck on 6nm which isn't all that much better than 7nm hence the very small improvements if any.

The 7900XT and XTX dies are 529sq.mm while the 4080 is 379sq.mm.
And the 4090 is 609...
How much bigger could AMD have made it?

The die size is misleading due to the MCD being 6nm..... I'd expect them to be able to do a GCD that's 20-30% larger though.

 

[gffermari]

The die size is misleading due to the MCD being 6nm..... I'd expect them to be able to do a GCD that's 20-30% larger though.

Oh ok.
On the other hand, I think that was the biggest die they could get for the GCD, so it can be within power spec, because separating the monolithic to 6+1 parts, there has to be some space for the connections.

 

[Tropick]

I'm still optimistic.... AMD first generation chiplet cpu had issues and have improved by leaps and bounds every generation. Maybe I am being naïve but I am hoping for a similar progression. I've said this 100 times but we need all 3 gpu vendors to be competitive or else progress will stagnate. Just like on the CPU side its best when both AMD and Intel offer compelling products at every price point.

This is kind of how I'm looking at it too. RDNA3 definitely feels like a beta test for chiplet-based GPUs, sort of reminds me of Zen 2. I have high hopes for RDNA4 but they really need to put forth significant effort to get ray tracing on the same level as NV. Not to mention the 4090 is just in a different league all together, no matter how you cut it.

Their drivers still need a lot of work too. They're still trying to sort out the high power usage on RDNA3 (albeit some issues have been fixed), you STILL can't do a clean install using the first party installer without risking bricking your OS, and Rift Apart not launching with any RT support was absolutely goofy. And that's just to name a few.

Back when I had a 980Ti, I seriously don't think I ever had a driver issue, ever. I love my 6950XT but it's been nearly 3 years since RDNA2's launch and it's only recently that I've been able to say that the drivers are "solid".

 

[oxrufiioxo]

Oh ok.
On the other hand, I think that was the biggest die they could get for the GCD, so it can be within power spec, because separating the monolithic to 6+1 parts, there has to be some space for the connections.

View attachment 307002

They obviously felt it wasn't economical but if Nvidia can release 450w gpu they should be able to also..... It would have at least been interesting I find it hard to believe they couldn't do a bigger GCD than 300mm2

 

[AnotherReader]

They obviously felt it wasn't economical but if Nvidia can release 450w gpu they should be able to also..... It would have at least been interesting I find it hard to believe they couldn't do a bigger GCD than 300mm2

It isn't a technical limitation rather an economical one, even if self-imposed.

 

[kapone32]

I have a different take on pricing. People expect that AMD is responsible for pricing. That however is not 100% true. Why are (were) All in One desktops so much cheaper. Well one of the reasons was the Distributor was cut out. In the World of GPUs where I live there are levels that must be crossed before a card comes into your hands. In Canada before I can buy a GPU it must be sold from the Manufacturer to the Distributor and then onto the retailer. Retailers have different agreements with distributors but with Mining and Nvidia the Distributors became more greedy for GPUs. As evidence one of the most maligned cards is the Asus Dual so the 6600 has been the cheapest capable GPU on Canada Computers site for a while (until they sold them all) at $269.99. It would seem that As Rock are selling directly to retailers though as their cards are always the cheapest. I have posted the 6700XT on Newegg and you will notice the As Rock board has been $459.99 CAD for months. Even the 6800XT As Rock variant has been $749 for months now but the 7900XT is the cheapest 7900XT you can buy on Newegg and you can go further down the stack and see that only the Asus Dual competes on pricing but those are usually the last run for a GPU and never come into the high space. I bought a 6750XT for a customer build for $419 ($319 US) and they are over the moon.

6700xt | Newegg.ca

Search Newegg.ca for 6700xt. Get fast shipping and top-rated customer service.

www.newegg.ca

The other issue is how great the 6000 series cards are. If you own a 6500XT you are amazed if you open the card and see the size of the GPU. If you have any card above a 6600 you are Golden as long as you play at the right resolution. Even though the narrative has people championing things like DLSS there is no reason to use those with AMD. They are not the evil empire and we will all be happy when we can build Mini ITX builds with APUs that support 1080P 120+ HZ on 4K Tvs for stellar Gaming. I have said this before but if you really look at it AMD is killing it right now. Even Sony is bringing back handheld and now we know that the Desktop APU will have Vcache as well. The release of 7000 was not the best but if you bought one of those GPUs you understand why there is not a lot of 7900s on used markets. When the 7800XT is launched it should bring things back into balance as that should be a $600 MSRP in this environment but could be $500 by December. We expect 100% improvement when that is not always viable> I read post stating that the 7000 series was only 35% faster but in reality going to 115 FPS from 85 can be felt at 4K especially to make Freesync even better. I play a ton of Total War and that Game will bring your PC to it's knees but I get a consistent over 100 FPS in Battles no matter the size of Battles at 4K. With the 6800XT I had to set the Game to 1440P to enjoy it.