AMD Radeon RX 6000 Series Specs Leak: RX 6900 XT, RX 6800 XT, RX 6700 Series

[Xmpere]

If the 6700 is in 250-350 range. Instant cop. Gonna be my first gpu if it is.

 

[Chrispy_]

Nonsense, there are excellent models like the Asus RTX 3080 TUF, unless you are going to overclock, which most buyers don't do.

How is that MSRP? it's £100 more than the FE on Nvidia's official 3080 page and £30 more than other entry-level cards from other AIB vendors.

https://www.nvidia.com/en-gb/shop/geforce/gpu/?page=1&limit=9&locale=en-gb&category=GPU&gpu=RTX%203080&sorting=lp

 

[TheoneandonlyMrK]

Well, if it's L1 cache, it's going to be tiny. And please note they narrow down the specific scenario where that cache helps.
Don't get me wrong, caching works, it improves performance. But it's different from bandwidth. I mean, look at Intel's CPUs. Those have had pretty smart caches (three levels of them) that do a pretty good job. But they still only hide latency, they're not a substitute for bandwidth.
Of course, GPU usage patterns are "a little" different, but I hope you get the idea.

GPU don't have 3 tier cache and think what your now saying!?.
They have one unified tier now and unifying that just helped zen 3 to a 19% improvement.
I get you and we'll see, it's not long but no need for the hard stances either.

 

[bug]

GPU don't have 3 tier cache and think what your now saying!?.
They have one unified tier now and unifying that just helped zen 3 to a 19% improvement.

Not sure what you're trying to say, but Zen3 retains the 3 levels of cache, only it doesn't segment L3 cache per CCX (because it doesn't have CCXs anymore). But that's not responsible for the whole 19% perf uplift.

I get you and we'll see, it's not long but no need for the hard stances either.

No hard stance here, just keeping expectations in check.

 

[Zach_01]

GPU don't have 3 tier cache and think what your now saying!?.
They have one unified tier now and unifying that just helped zen 3 to a 19% improvement.
I get you and we'll see, it's not long but no need for the hard stances either.

To be fair, the +19% IPC gains was not only from unifing cache pool. ZEN3 has a lot more architectural changes/improvements.
Still we cant compare the 2 different chips (CPU/GPU). This is a first for a GPU and we have no baseline for what AMD cooked up with RDNA2.

 

[Vya Domus]

LOL, the 3090 is not a TITAN, nVidia point at it like if it was but it's not.

That's the point. Do you not find it strange the way they awkwardly said it's "TItan performance". Yeah, because that was meant to be a Titan.

 

[MxPhenom 216]

Can someone explain the difference between the 8nm and 7nm? Does that actually make a difference in performance or anything important?

Well TSMCs 7nm is way better than Samsungs 8nm. Samsungs 8nm is super old and its just an extension of their 10nm which isn't very exciting. Its got a bit better density and a higher drive cell option.

 

[TheoneandonlyMrK]

Not sure what you're trying to say, but Zen3 retains the 3 levels of cache, only it doesn't segment L3 cache per CCX (because it doesn't have CCXs anymore). But that's not responsible for the whole 19% perf uplift.

No hard stance here, just keeping expectations in check.

What I'm saying is look how your minimising what they did with zen3's cache and how it helped IPC and consider a GPU cache closer to it's cores!?, They're is a chance it could be good they showed as much with the Ryzen architecture, was always about cores,cache and connection.

 

[Chrispy_]

That's the point. Do you not find it strange the way they awkwardly said it's "TItan performance". Yeah, because that was meant to be a Titan.

People who bought Titans were buying them for the 24GB. We have a whole bunch of 2080Ti cards for rendering performance, but we had to buy a few Titans to cover those occasional workloads that don't run on 11GB cards. Titans are a ripoff, but they're still half the price of Quadros.

Big Navi is going to be interesting because the vanilla 6800 is a 16GB card for a fraction of the cost of a 2080Ti, a 3090, a Titan, or a Quadro RTX.
If you work in a cinematic / simulation industry then the raw performance isn't actually the key metric, it's whether the card in question can actually load the model(s) into VRAM or not. There's a metric I go by for certain types of purchasing that's not commonly useful, and that's $/GB for a GPU.

 

[kiddagoat]

I haven't posted regularly around here in quite some time. Why does it seem now, much more than before, that any AMD, Nvidia, or Intel article quickly deviates from the topic of the article and dives into back and forth fanboy loyalism???

This article is about AMD's upcoming product stack and not even a third of the way through here comes all the 30x0 series crap that's been regurgitated since the launch of Ampere. Maybe I am just old and out of touch. Why can't these discussions stay on topic?

 

[Chrispy_]

Why can't these discussions stay on topic?

Because Astroturf and Fanboy; Welcome to the 21st century.

 

[AsRock]

That's the point. Do you not find it strange the way they awkwardly said it's "TItan performance". Yeah, because that was meant to be a Titan.

It don't and really think it will have double the ram when they can do it. First of the 3090 is too cheap to be a Titen as it's not really for gamers.

 

[kostaspyrkas]

Performance isn't always the full picture, stable drivers also matters, look how the RX 5000 series launched, terrible drivers for months.

I have a 5700xt sapphire pulse almost from the beginning. .. no problems at all... and i play a wide range of games

 

[Deleted member 185088]

Nvidia released a card 30% faster than a $1200 card for $1099 and they expected to sell only 5 cards. Lol

Fixed that for you, how come Turing was priced the same everywhere (except where taxes are the issue), but Ampere costs up to 40% outside the US.

 

[Ibotibo01]

RDNA2 will aged like wine. GTX 600 series'(even 700 series) performance are worse than HD 7000 series. HD 7970 gives same performance with GTX 780/Ti in 2020/2019 games.

 

[bug]

What I'm saying is look how your minimising what they did with zen3's cache and how it helped IPC and consider a GPU cache closer to it's cores!?, They're is a chance it could be good they showed as much with the Ryzen architecture, was always about cores,cache and connection.

And what I'm saying is you don't seem to understand what cache does and how it works

 

[Flanker]

I haven't posted regularly around here in quite some time. Why does it seem now, much more than before, that any AMD, Nvidia, or Intel article quickly deviates from the topic of the article and dives into back and forth fanboy loyalism???

This article is about AMD's upcoming product stack and not even a third of the way through here comes all the 30x0 series crap that's been regurgitated since the launch of Ampere. Maybe I am just old and out of touch. Why can't these discussions stay on topic?

It's been like that for almost every forum I visited in the last 15 years

 

[Shou Miko]

With the current shortages of Nvidia's RTX 3080 and specially cards like the Asus Tuf model I am still thinking about going AMD this time around.

I am only sad to see that the RX 6900 XT (Navi 21 XTX) is a AMD exclusive part like Nvidia's Titan cards are wish AIB partner would get clearance to make custom versions of the Navi21 XTX models.

But I am still hoping even it will be a ref design I can get it with Sapphire branding so I get keep using Sapphire's TriXX software because of the scaling option that I use on my 4K screen from time to time.

 

[Vya Domus]

I am only sad to see that the RX 6900 XT (Navi 21 XTX) is a AMD exclusive

I am not sure what you mean by exclusive, it's said that AMD will only allow reference models for the 6900XT not that they will only be sold through AMD.

 

[Shou Miko]

I am not sure what you mean by exclusive, it's said that AMD will only allow reference models for the 6900XT not that they will only be sold through AMD.

With exclusive I mean no custom PCB and cooling solutions but what about the bios? that will properly show AMD as vendor so I don't believe that the Sapphire TriXX software will then work.

 

[TheoneandonlyMrK]

And what I'm saying is you don't seem to understand what cache does and how it works

Yeh yeah, some are just less negative.

 

[ShurikN]

no custom PCB

AMD's reference PCB and board design are one of the best on the market. Everything is over-engineered, and it rivals much more expensive Sapphire models for example.

 

[renz496]

28th is the official unveil. when can we expect actual review?

 

[FinneousPJ]

28th is the official unveil. when can we expect actual review?

Based on Zen 3, on launch day.

 

[bug]

28th is the official unveil. when can we expect actual review?

Upcoming Hardware Launches 2023 (Updated Feb 2024)

This article serves as a continuously updated summary of currently known leaks and official announcements regarding upcoming hardware releases in 2024 and beyond. We cover and keep track of developments for Intel Arrow Lake, AMD Zen 5, NVIDIA's and AMD's upcoming GPUs, DDR6 and GDDR7 memory...

www.techpowerup.com