ASRock Radeon RX 7800 XT Spotted in EEC Registry (listed with 16GB)

Published by

Click here to post a comment for ASRock Radeon RX 7800 XT Spotted in EEC Registry (listed with 16GB) on our message forum
https://forums.guru3d.com/data/avatars/m/255/255510.jpg
Sounds good, bet the development team guys at AMD are buzzin on their endeavours. Then Lisa says" Hey, it actually works, so lets charge stupid prices for it".
https://forums.guru3d.com/data/avatars/m/248/248627.jpg
I always assumed this would use the same chiplets from the 7900xtx @ 1024 shaders per die with it having 4 dies for 4096 shaders guess we'll wait and see, i dont see too many ppl talk about the 7900xtx some claim the mcm tech isn't ready yet
https://forums.guru3d.com/data/avatars/m/271/271560.jpg
icedman:

I always assumed this would use the same chiplets from the 7900xtx @ 1024 shaders per die with it having 4 dies for 4096 shaders guess we'll wait and see, i dont see too many ppl talk about the 7900xtx some claim the mcm tech isn't ready yet
the people that talk about the 7900xtx mcm design negatively haven't played on one. it's a stone cold killer (in a good way) the talk about inter-module latency is the same as debating the number of angels that would fit on a pin. it was a historic issue in the development of mcm gpu technology. the simple fact that the more expensive rtx 4080 (when not memory bound) trades blows with it in every performance chart is the evidence. if you're buying a gpu for RT, you either play one of a handful (70 or so) of games or are a sucker for slick and glib marketing. if you're buying a (high end) gpu for dlss - save money and buy a cheaper card
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
This is the gpu im looking as my upgrade path. I guess amd wont release it until most 6000 stock is gone.
https://forums.guru3d.com/data/avatars/m/271/271560.jpg
Undying:

This is the gpu im looking as my upgrade path. I guess amd wont release it until most 6000 stock is gone.
that was the plan until now. but the market is too soft and warehousing is expensive in lost oppportunity. so voila! if ppl thought selling rdna 2 was getting old, the midrange rdna 3 is aging as we speak and Nvidia is happy to fill the gaps
https://forums.guru3d.com/data/avatars/m/282/282473.jpg
icedman:

I always assumed this would use the same chiplets from the 7900xtx @ 1024 shaders per die with it having 4 dies for 4096 shaders guess we'll wait and see, i dont see too many ppl talk about the 7900xtx some claim the mcm tech isn't ready yet
What shader dies ? There is one graphics die on n31. Chiplets are for cache only.
https://forums.guru3d.com/data/avatars/m/248/248627.jpg
cucaulay malkin:

What shader dies ? There is one graphics die on n31. Chiplets are for cache only.
You are correct, I had mistaken the dies I thought the 6 dies were the shaders but it was the other way around, main core with cache dies added. Hadn't really looked too much into just assumed with amd going multi die they would be splitting the shaders up.
https://forums.guru3d.com/data/avatars/m/282/282473.jpg
icedman:

You are correct, I had mistaken the dies I thought the 6 dies were the shaders but it was the other way around, main core with cache dies added. Hadn't really looked too much into just assumed with amd going multi die they would be splitting the shaders up.
it's just memory that's moved to the side of the die (which also increases latency a lot). https://www.techpowerup.com/309686/amd-radeon-rx-7600-gpu-has-better-cache-vram-latency-than-rx-7900-xtx
Chips and Cheese reports that AMD's RX 7900 XTX takes up to 58% longer to access and pull data from its pool of Infinity Cache, when contrasted with the recently released sibling. The difference is especially large with RDNA 3. With RDNA 2, the RX 6900 XT had 151.57 ns of Infinity Cache latency compared to 130 ns on the RX 6600 XT, or a 16.5% latency penalty for the larger GPU.
https://forums.guru3d.com/data/avatars/m/275/275921.jpg
Wait for the biggest Navi.
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
AuerX:

Wait for the biggest Navi.
The biggest Navi is what you can afford.
https://forums.guru3d.com/data/avatars/m/282/282473.jpg
AuerX:

Wait for the biggest Navi.
what's coming that's bigger than 7900xtx ? and will it run on one wall socket ?
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
cucaulay malkin:

what's coming that's bigger than 7900xtx ? and will it run on one wall socket ?
Maybe it will start burning like the 4090's 😀
https://forums.guru3d.com/data/avatars/m/282/282473.jpg
Undying:

Maybe it will start burning like the 4090's 😀
more likely to use 5x8 -pin.
https://forums.guru3d.com/data/avatars/m/271/271560.jpg
cucaulay malkin:

what's coming that's bigger than 7900xtx ? and will it run on one wall socket ?
hilarious 😛 ...the 7950xtx (but idk if they're still going to launch it...economics) btw the MI1300 has a draw of 750w - but it is more efficient than 7900xtx and makes the 7900xtx look weak https://cdn.mos.cms.futurecdn.net/HrphTm7vXbSbNp5L3cNcHm-970-80.jpg.webp
https://forums.guru3d.com/data/avatars/m/248/248627.jpg
^I seen that they had another version on techpowerup and from what i can tell the only differences are more cache per die (192mb vs 96mb) and a 450w tdp allowing for 3300mhz which is quite significant.
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
AMD fucked up the the naming. 7900xt should be 7800xt and 7900xtx should be 7900xt. Then you can release navi32 7700xt with navi21 6900xt performance price it 499$ and be done with it. Nvidia played a role there with renaming the 4080 12gb to 4070ti i think. There is a picture on the videocardz navi32 collecting dust they just cant release it with such a miniscule uplift and pricing they want. Its sad, the whole rdna3 should be skipped and start working on rdna4.
https://forums.guru3d.com/data/avatars/m/216/216349.jpg
Undying:

AMD ****** up the the naming. 7900xt should be 7800xt and 7900xtx should be 7900xt. Then you can release navi32 7700xt with navi21 6900xt performance price it 499$ and be done with it. Nvidia played a role there will renaming the 4080 12gb to 4070ti i think. There is a picture on the videocardz navi32 collecting dust they just cant release it with such a miniscule uplift and pricing they want. Its sad, the whole rdna3 should be skipped and start working on rdna4.
AMD didn`t mess up the naming, they are doing it on purpose so they can ask more for their GPUs. AMD probably regrets selling the 6800XT at 649€ because we are talking about a very big chip that was sold with a minimal margin in the 6800XT form. Now to correct this "mistake", they renamed the true sucessor of the 6800XT as the 7900XT, moving a tier up and asking more for it. And they are going to replace the 6800XT with a much smaller chip and sell it at the same price, increasing their margins considerably. I don`t like this move but it makes sense from a business perspective...
https://forums.guru3d.com/data/avatars/m/271/271560.jpg
cucaulay malkin:

Same as h100 iirc
ahh nope. this monster has 24 Zen 4 cores,CDNA gpu cores, & 128Gb HBM3... one blade of El Capitan (2 Exaflops) has 4 watercooled MI 1300 in a 1u case
https://forums.guru3d.com/data/avatars/m/282/282473.jpg
H83:

AMD didn`t mess up the naming, they are doing it on purpose so they can ask more for their GPUs. AMD probably regrets selling the 6800XT at 649€ because we are talking about a very big chip that was sold with a minimal margin in the 6800XT form. Now to correct this "mistake", they renamed the true sucessor of the 6800XT as the 7900XT, moving a tier up and asking more for it. And they are going to replace the 6800XT with a much smaller chip and sell it at the same price, increasing their margins considerably. I don`t like this move but it makes sense from a business perspective...
Makes less sense now since rdna 3 has not delivered much gen to gen performance uplift, a smaller 7800xt would mean near last gen performance for the same price tier. That is why navi 21 owners will be really disappointed with 7800xt. A Rdna2 moment likely won't happen again especially on expensive node with architecture that failed to utilize dual issue fp32 in games properly and fell short of their own presentation goals even.