AMD Radeon RX 8000 RDNA 4 GPUs rumored to use slower 18 Gbps GDDR6 memory

DragonSlayer101

Posts: 377   +2
Staff
Rumor mill: AMD is expected to launch its Radeon RX 8000 GPU lineup later this year, and while it may not include high-end models, mainstream gamers are still buzzing with excitement over RDNA 4. However, some may be disappointed with a new rumor suggesting Team Red could use slower GDDR6 memory in its next-gen GPUs.

The rumor comes from generally reliable tipster @Kepler_L2, who claims that the Radeon RX 8000 GPUs will ship with 18 Gbps GDDR6 memory instead of newer solutions like Samsung's GDDR6W DRAM, which doubles the performance and capacity of GDDR6.

If the claim turns out to be accurate, it will be a highly surprising decision from AMD, especially since Nvidia is expected to launch its RTX 5000-series cards later this year with 28Gbps GDDR7 memory.

AMD has been using 18Gbps memory in many of its current-generation graphics cards like the RX 7900 GRE and RX 7700 XT, as well as all the Navi 33 models, ranging from the Radeon Pro W7600 to the Radeon RX 7600 XT. In fact, AMD's current premium offerings, such as the 7900 XTX and the 7900 XT feature faster memory, with speeds of up to 20Gbps. Even the RX 7800 XT comes with 19.5Gbps memory.

AMD's use of 18 Gbps GDDR6 could be a major disappointment for gamers hoping for faster memory in the upcoming cards. That said, the RX 8000 series is expected to cater to the mainstream and entry-level markets, so the inclusion of the older technology could be a way for AMD to keep the prices low.

It could also help keep the power consumption in check, as the next-gen solutions are said to consume more power. Either way, all this is still conjecture, as AMD is yet to officially clarify its plans for RX 8000.

While the Radeon RX 8000 line was initially expected to feature high-end cards for enthusiasts and serious gamers, more recent rumors suggest that it will only include mid-range and entry-level models for the mainstream market.

The two expected SKUs in the RX 8000 series are the Navi 48 and Navi 44 that were spotted in ROCm patches last year. The former is expected to power the mid-range cards, while the latter is said to be the entry-level option.

Permalink to story:

 
RDNA4 should be very efficient if it can achieve 7900XT performance even with the restriction of lower bandwidth. It should bring good improvements to iGPUs which also have very limited bandwidths
 
This isn't really an issue if they don't limit the bus width. And if the the rumor of 7900xt performance for $400 is true, I don't care what their midrange card is, I'd say it'll be a mid ranged winner which is the largest market. It's not like 90% of the market are running 4090s. Also, we're really close to an economic collapse so the idea that high end cards will be a large market is laughable.

Using this slower memory might be a cost cutting measure. Frankly, I don't care about the high end. In a world of AAAA games that cost $130 where gameplay sucks and they cherry pick footage to make it look fun, I'll pass on investing thousands into hardware to flash pixels into my eyes.

Edit: FYI, I make tons of typos because I'm often posting from my phone. I'm not as dumb as my posts make me look, I'm actually much dumber than that in person
 
Last edited:
Ah the new talking points have arrived.. 'efficient', 'mid ranged winner' *scribbles furiously*
 
1, 2 and 3 grains of salt, next rumor please.
Whatever new gen GPU they are going to release, I'll wait till December to pull the trigger.
 
Hopefully the slower speed can be compensated for by other means, for example larger bandwidth or VRAM capacity, capable performance, lower power, and, of couse, lower price..
If it can, then it could be a winner for me..
is that too much compensation..?
 
Rumor mill: AMD is expected to launch its Radeon RX 8000 GPU lineup later this year, and while it may not include high-end models, mainstream gamers are still buzzing with excitement over RDNA 4. However, some may be disappointed with a new rumor suggesting Team Red could use slower GDDR6 memory in its next-gen GPUs.

The rumor comes from generally reliable tipster @Kepler_L2, who claims that the Radeon RX 8000 GPUs will ship with 18 Gbps GDDR6 memory instead of newer solutions like Samsung's GDDR6W DRAM, which doubles the performance and capacity of GDDR6. If the claim turns out to be accurate, it will be a highly surprising decision from AMD, especially since Nvidia is expected to launch its RTX 5000-series cards later this year with 28Gbps GDDR7 memory.

AMD has been using 18Gbps memory in many of its current-generation graphics cards like the RX 7900 GRE and RX 7700 XT, as well as all the Navi 33 models, ranging from the Radeon PRO W7600 to the Radeon RX 7600 XT. In fact, AMD's current premium offerings, such as the 7900 XTX and the 7900 XT feature faster memory, with speeds of up to 20Gbps. Even the RX 7800 XT comes with 19.5Gbps memory.

AMD's use of 18 Gbps GDDR6 could be a major disappointment for gamers hoping for faster memory in the upcoming cards. That said, the RX 8000 series is expected to cater to the mainstream and entry-level markets, so the inclusion of the older technology could be a way for AMD to keep the prices low. It could also help keep the power consumption in check, as the next-gen solutions are said to consume more power. Either way, all this is still conjecture, as AMD is yet to officially clarify its plans for RX 8000.

While the Radeon RX 8000 line was initially expected to feature high-end cards for enthusiasts and serious gamers, more recent rumors suggest that it will only include mid-range and entry-level models for the mainstream market. The two expected SKUs in the RX 8000 series are the Navi 48 and Navi 44 that were spotted in ROCm Patches last year. The former is expected to power the mid-range cards, while the latter is said to be the entry-level option.

Permalink to story:

Interesting rumor about AMD using 18 Gbps GDDR6 in the RX 8000 series. If true, it could be a cost-saving measure to target mainstream gamers, especially considering Nvidia's rumored faster memory for the RTX 5000.
Still, hopefully the focus on mid-range and entry-level translates to good value for those looking to upgrade. We'll have to wait for official info from AMD to know for sure.
 
Long as it’s 7900 xt level of performance for >$499 and the entry level is 7800 xt for >349$ while using 100+ less watts would be great. 215w 8700xt and 160w 8600xt INC!
 
Ah well, bar keeping something in the tank for an RDNA5 return at least to current form, placing and pricing, it was a damn good try at a comeback for AMD.

Again, as said before... I had a damn good run with a 6800XT and 7900XTX, together something like a two grand saving vs the 3080 10Gb and 4080 alternatives for, let's be totally honest here, not far less performance where it counts the most. I am actually bummed about this but it's unsurprising and I can't blame AMD at all. Whatever other reasons may be given one of the major issues is that they likely couldn't sustain and afford their current pace of progress vs the significant lack of interest and purchases for the comparative value they offered the last two gens, especially at the top 2-3 tiers.

Now we face a scenario where AMD are holding back, Intel's Battlemage will most likely be only as good at best and the top 2-3 tiers will be absolutely ruled by Nvidia... who, along with partners, I doubt will slow down on the insane pricing. Add in the usual extra cost of retail over MSRP increasing price range both per tier and within a tier (along with a possible conveniently short supply and availability) and I can see record prices again by some hundreds more even for the less desirable near base/ref cards. Same old story then, only this time there won't be any remotely near par, cheaper alternative.

The only winners here are Nvidia, their shareholders and game devs, who'll have a nice new gen restricted version of DLSS to easily replace the work of optimisation again. Honestly, at this rate my next 'upgrade' will come with a resolution downgrade or greater focus on consoles... though the latter lacks the games and genres I love the most and bother going this far to maintain a PC for.
 
FWIW:
AMD is releasing the RDNA4 cards for the ~$600 & under crowd. So essentially the performance of the 7900xtx using less energy and costing less money, using one die.

AMD is also releasing RDNA5 this year... which is a multi-chip design like the 7000 series flagship, but they will be Custom designed cards with multiple configurations (ie multi-chip) to suit your gaming desire/needs ($999 ~$2,999k) for 4k+ and Pro gaming needs.


CUDA is dead for gaming, NVidia already admitted this and pushing the Ai wave, instead of engineering new gaming architecture like RDNA is. I use to buy EVGA only, now 1 Year with my XTX jamming Guildmates who went 4090 on Call of Duty, makes my point. They are all awaiting RDNA5 for their new displays.

 
AMD is also releasing RDNA5 this year... which is a multi-chip design like the 7000 series flagship, but they will be Custom designed cards with multiple configurations (ie multi-chip) to suit your gaming desire/needs ($999 ~$2,999k) for 4k+ and Pro gaming needs.
Can I buy pot from you?
 
The question is whether this is going to materially result in lower performance? I think a significant jump in memory bandwidth may result in some performance lost, but materially, I am not too sure. A good example now is the RTX 4070 Ti versus 4070 TI Super where the memory bandwidth increased materially, but performance gain is just in the low single digit %. At the end of the day, if the GPU couples these GDDR6 with 192/256 bit memory bus, I think we are not losing that much of a memory bandwidth. And I feel there may be some trade offs in the form of latency spikes as these memory gets faster.
 
FWIW:
AMD is releasing the RDNA4 cards for the ~$600 & under crowd. So essentially the performance of the 7900xtx using less energy and costing less money, using one die.

AMD is also releasing RDNA5 this year... which is a multi-chip design like the 7000 series flagship, but they will be Custom designed cards with multiple configurations (ie multi-chip) to suit your gaming desire/needs ($999 ~$2,999k) for 4k+ and Pro gaming needs.


CUDA is dead for gaming, NVidia already admitted this and pushing the Ai wave, instead of engineering new gaming architecture like RDNA is. I use to buy EVGA only, now 1 Year with my XTX jamming Guildmates who went 4090 on Call of Duty, makes my point. They are all awaiting RDNA5 for their new displays.
RDNA5 is not yet ready. It won't come out this year. Probably summer 2025 at the earliest.
 
Obviously they won't be using expensive GDDR7 on low to mid-end cards which Radeon 8000 is going to be.

The best RDNA4 SKU will be as fast as 7900XT at most.

RDNA5 is very late 2025 or even 2026, they are not even close and 5090 and 5080 comes out in about 6 months.

Funny how Nvidia can dominate gaming market while having 90% focus on AI.

AMD soon drops below 10% dGPU marketshare on Steam HW Survey. That is just reality and AMD GPU fanboys can deny it all they want. No-one cares.

AMD makes great CPUs - I use one - but GPUs, nah. Lacking in too many areas. AMD GPUs are cheaper for a reason and still don't sell. Mostly because of wonky drivers, mediocre features and lacking RT perf + Resell value is low + low demand as well. Higher power draw. You save nothing in the end.
 
Last edited:
Still too much but this is exactly why resell value is low on AMD hardware. Constant pricedrops. Together with low demand, since 9 out of 10 want Nvidia.

A friend of mine bought 7900XTX on release for 1150, he is currently crashing left and right in Helldivers 2 like tons of AMD GPU owners. Funny considering its a PS5 port made for AMD hardware in the first place. He is waiting for 5080 and will switch instantly.

Just shows that 90% of PC game developers focus on Nvidia. Nvidia just works way better overall, especially in lesser popular titles, early access games and emulators. I do it all.

AMD focus on popular games that gets benchmarked alot for reviews and nothing else. They very often lack launch drivers for new games. Nvidia always have day one drivers ready and most game dev's test games using Nvidia hardware as well.
 
Still too much but this is exactly why resell value is low on AMD hardware. Constant pricedrops.

A friend of mine bought 7900XTX on release for 1150, he is currently crashing left and right in Helldivers 2 like tons of AMD GPU owners. Funny considering its a PS5 port made for AMD hardware in the first place.
What about 4090 suprim liquid at $1669

or Check this out on @Newegg:GIGABYTE GeForce RTX 4090 WINDFORCE V2 24G Graphics Card, 3x WINDFORCE Fans, 24GB 384-bit GDDR6X, GV-N4090WF3V2-24GD Video Card https://www.newegg.com/gigabyte-gef...m_mmc=snc-social-_-sr-_-14-932-625-_-04242024

Just like I predicted the market rejected 4090s >2k usd.
 
What about 4090 suprim liquid at $1669

or Check this out on @Newegg:GIGABYTE GeForce RTX 4090 WINDFORCE V2 24G Graphics Card, 3x WINDFORCE Fans, 24GB 384-bit GDDR6X, GV-N4090WF3V2-24GD Video Card https://www.newegg.com/gigabyte-gef...m_mmc=snc-social-_-sr-_-14-932-625-_-04242024

Just like I predicted the market rejected 4090s >2k usd.

The price for 4090 went up because of massive demand. I paid 1500 dollars on day one, and will be able to sell it for 1000 dollars when I upgrade to 5090 in ~6 months.

500 dollars "lost" for 2 years with top tier performance is great. Try that with an AMD card ;)
AMD won't even have 4090 performance before RDNA5 in 2026.

AMD is not competiting in high-end anymore.

The sole purpose of RDNA4 is to regrab marketshare, because they lost tons with 7000 series.


Sad but true. Lets hope AMD can win some back with RDNA4.
 
Back