Friday, August 25th 2023

AMD Unveils Radeon RX 7800 XT and RX 7700 XT Graphics Cards

AMD today at Gamescom unveiled the Radeon RX 7800 XT and Radeon RX 7700 XT performance-segment graphics cards. Designed for maxed out gaming at 1440p with ray tracing, the two are designed to square off against NVIDIA's GeForce RTX 4070 series, offering competitive performance and pricing. The two are based on AMD's latest RDNA3 graphics architecture, and use the 5 nm foundry process where it matters. Both cards claim to offer not just superior performance to the specific NVIDIA RTX 40-series SKUs they're designed to compete with, but also better future-proofing, with more video memory on offer.

At the heart of the two is the new "Navi 32" GPU, AMD's second largest chip from this generation. It is a chiplet GPU, just like the "Navi 31" that powers the RX 7900 series, albeit slightly scaled down. The graphics compute die (GCD), the die with the main graphics rendering and compute machinery, is built on the 5 nm EUV foundry node. It is flanked by four memory cache dies (MCDs), each built on the 6 nm foundry node. These are the same MCDs found in the "Navi 31," but four in number instead of six, which gives the "Navi 32" a 256-bit wide GDDR6 memory interface.
The Radeon RX 7800 XT maxes out the "Navi 32," enabling all 60 compute units (CU) physically present, which works out to 3,840 stream processors, 120 AI accelerators, 60 Ray accelerators, 64 MB of Infinity Cache, and 16 GB of GDDR6 memory across the chip's full 256-bit memory interface. The GPU ticks at 2124 MHz Game clocks, and 2430 MHz boost; while its memory runs at 19.5 Gbps, resulting in 624 GB/s of memory bandwidth. The card is configured with 263 W of total board power, and its reference design comes with two 8-pin PCIe power connectors.
The Radeon RX 7700 XT is cut down from the same "Navi 32" silicon as the RX 7800 XT, and uses the same reference board design. The biggest change here, is that the memory size is reduced to 12 GB, the Infinity Cache to 48 MB, and the memory bus width to 192-bit. One of the four MCDs on the "Navi 32" silicon is disabled. Over on the GCD, the RX 7700 XT is configured with 54 CU, which works out to 3,456 stream processors, 108 AI accelerators, 54 Ray accelerators, and 180 TMUs. The GPU runs at higher clock speeds than the RX 7800 XT, with 2171 MHz Game clocks, and 2544 MHz boost. The memory speed, however, is lower, at 18 Gbps, which over the 192-bit memory interface, puts out 432 GB/s of bandwidth. The RX 7700 XT is configured with 245 W of total board power, and has the same dual 8-pin power input setup as the RX 7800 XT.
In terms of performance, AMD claims that the Radeon RX 7800 XT offers anywhere between 2% to 23% performance gains over the GeForce RTX 4070 in 13 of the 19 games they tested. Testing was done at 1440p with max settings for each game. AMD also claims that the extra 4 GB of memory should give you better future-proofing. The Radeon RX 7700 XT, on the other hand, is shown scoring anywhere between 1% and 31% higher than the GeForce RTX 4060 Ti 16 GB, in 16 out of 19 games that the two cards were compared in.

The reason AMD chose the RTX 4070 and RTX 4060 Ti 16 GB for comparisons, is because it intends to price the RX 7800 XT and RX 7700 XT competitively to them. The Radeon RX 7800 XT is priced at USD $499, while the RX 7700 XT is priced at $449. Both cards go on sale from September 6, 2023.
Add your own comment

104 Comments on AMD Unveils Radeon RX 7800 XT and RX 7700 XT Graphics Cards

#76
TumbleGeorge
Beginner Micro Deviceespecially if FSR3 won't be AS good
Yes "if". This is the problem, because you suppose something without real information about it.
Well after all is possible that you is right.
Do you have any inner information how good or bad is FSR3?
Posted on Reply
#77
kapone32
Beginner Micro DeviceI don't predict anything for FSR3. I'm predicting the fate of lower end RDNA3 which is doomed, especially if FSR3 won't be AS good. AMD is habitually a generation behind, today they're expanding to being two generations behind (DLSS3.5 is already here but FSR3 is nowhere to be seen; ray tracing in nVidia GPUs improved two times already but AMD products stay on the same level as they were back in 2020). Yeah, raw raster performance is on par. This is good if we cut it outta context. This is a complete disaster if we get all the pieces together.

RDNA3 pros:
+ Higher VRAM compared to competition
+ Some AMD exclusive features
+ More raw raster performance per dollar
RDNA3 cons:
- EVERYTHING ELSE.
Let me see I Game as a hobby with my PC and the biggest Game right now is Baldur's Gate 3. So let's see

Baldur's Gate 3 Good

VRAM
Raw performance
Adrenline Software support

Everything else: Fluff

I love how people make raster seems like it is not the foundation of PC Gaming.
Posted on Reply
#78
Beginner Macro Device
TumbleGeorgeDo you have any inner information how good or bad is FSR3?
Of course I do. It's 0 outta 10 because it's not released yet. This is the ONLY thing matters right now. AMD increases their failure by delaying the only thing capable of giving some life to their everdead product line-up.

And of course I don't because I'm just a user and not some FBI guy who has insider information on the product.

One. More. Time. The only thing relevant, once again, RELEVANT is RDNA3 bad performance per dollar and bad feature arsenal per dollar. Extremely weak RT, very weak energy efficiency, stone age scaling tech, no CUDA, hardly limited overclocking et cetera, et cetera. FSR3 is just a crutch which could've helped this sorry horsepoop of a generation but it's yet to be seen. And it has to be superior to DLSS3 to actually do so. Which, knowing AMD, is less likely than a deployed nuke in my bathroom.
kapone32I love how people make raster seems like it is not the foundation of PC Gaming.
I love how people prove their point only using one game despite billions of them existing. This is called nitpicking and it's a fallacy.
Posted on Reply
#79
TumbleGeorge
Beginner Micro DeviceAMD increases their failure
What you mean as failure? AMD have big HPC and AI market progress. Demonstrated huge today and future interest for their Instinct MI300&MI300X from their main costumers, which prompted the need for increased production and additional orders for components. For example, the recently announced order to Samsung for HBM3 memories. Who knows, the gaming graphics card sector might still be a bit more important to AMD than to Nvidia. But whatever happens in it, even when it's a problem for the average Joe, is unlikely to actually hurt AMD's business success.
Posted on Reply
#80
Beginner Macro Device
TumbleGeorgethe gaming graphics card sector
TumbleGeorgeWhat you mean as failure?
This is what I mean. I know they are doing well in other fields but this one, they are just throwing abandonware at us. They will definitely lose this market segment if RDNA4 will be the same as RDNA3 in terms of generational uplift (meaning 10 percent sharp).

RDNA2 wasn't better than Ampere. Neither is RDNA3. But it's already Ada out there and also Intel pondering upon them. I know Intel's position is currently very weak but they might surprise us. They might not. Who knows. But they're definitely not to be underestimated.

All this just feels like they're not interested in gaming anymore.
Posted on Reply
#81
rv8000
TaishoIt's been 3 years. If AMD didn't fail with Navi 3 big time, they would have no problem comparing 7800 XT to 6800 XT.
Explain to all of us how big a failure navi 3 is then. And as a bonus tell us using the same wisdom that anything but a 4090 isnt a failure for the exact same reasons.
Posted on Reply
#82
kapone32
Beginner Micro DeviceOf course I do. It's 0 outta 10 because it's not released yet. This is the ONLY thing matters right now. AMD increases their failure by delaying the only thing capable of giving some life to their everdead product line-up.

And of course I don't because I'm just a user and not some FBI guy who has insider information on the product.

One. More. Time. The only thing relevant, once again, RELEVANT is RDNA3 bad performance per dollar and bad feature arsenal per dollar. Extremely weak RT, very weak energy efficiency, stone age scaling tech, no CUDA, hardly limited overclocking et cetera, et cetera. FSR3 is just a crutch which could've helped this sorry horsepoop of a generation but it's yet to be seen. And it has to be superior to DLSS3 to actually do so. Which, knowing AMD, is less likely than a deployed nuke in my bathroom.

I love how people prove their point only using one game despite billions of them existing. This is called nitpicking and it's a fallacy.
WTF what one Game do you think that in my 800 Game library that BG3 is the only one I play and am not the only one I enjoy.
Posted on Reply
#83
ARF
Beginner Micro DeviceRDNA2 wasn't better than Ampere.
Navi 21 is 520 sq. mm vs. GA102 628 sq. mm. That is 20-21% larger die. Vs-vs Radeon RX 6950 XT is only 16% slower than RTX 3090 Ti which actually means that RDNA2 was better than Ampere..
Also, RX 6800 XT was launched with MSRP of only 650$ against later launched RTX 3090 Ti for 2000$ for only 35ish% higher performance.

AMD was competitive with RDNA2, today it's not with RDNA3 mainly because of the greedy pricing of both RX 7900 XT and RX 7900 XT which should have been RX 7800 XT and RX 7800 :D
Posted on Reply
#84
Assimilator
kapone32The only cards that come with USB C are Reference designs. I have never seen an AIB card with USB C.
And the image shown in the OP is a reference card. With no USB-C.
Posted on Reply
#85
N/A
ARFNavi 21 is 520 sq. mm vs. GA102 628 sq. mm. That is 20-21% larger die. Vs-vs Radeon RX 6950 XT is only 16% slower than RTX 3090 Ti which actually means that RDNA2 was better than Ampere..
ONLY because of 8NM vs N7 and they lost that advantage. now 7900 vs 4070 Ti is 300 and 36,6x5 in chiplets vs 300mm monolithic and it's a role reversal. 384 bit vs 192 bit. was 256 vs 384 lol.
and it is 9% if you look at the real test average though. depending on the title it is more like +- double digit. all over the place.
Posted on Reply
#86
tfdsaf
7700XT needs to be $350 and they also need another card to slot in at $300, the 7600 is realistically an entry level $220 card at most!
Posted on Reply
#87
Taisho
rv8000Explain to all of us how big a failure navi 3 is then. And as a bonus tell us using the same wisdom that anything but a 4090 isnt a failure for the exact same reasons.
Here you go:
- AMD is not able to create a product that can compete in raster performance with 4090.
- Claiming "Efficiency leadership" when they already know how efficient Ada is, a couple of weeks ahead of its sales, is not a manipulation but a defiant lie.
- Marketing low manufacturing costs of multi-chip modules, and then launching their GPUs with prices resembling crypto boom launch prices.
- The release is plagued with typical AMD problems like high multi-monitor power draw that based on forum posts is still not fully fixed.
- Vsync power consumption until the monolithic 7600 is almost 2 times higher vs. Ada and worse than Ampere.
- AMD manages to mess up the video playback power draw in 7600.
- Calling 100+ Celsius junction temps on max fan RPM for their reference models working as intended until the huge outrage from the community.
- The 7900 XT is just an upsell product, so basically their lineup is only 2 GPUs: 7900 XTX, which at this price point makes no sense versus RTX 4090, and RX 7600 priced too close to the faster RTX 4060, with far worse value than Navi 2 options.

The only problems with NVIDIA GPUs this gen are prices and carrying on with the plan to make VRAM a luxury good. AMD was quite happy to follow the suit though, locking >8GB RAM behind a solid paywall.
Posted on Reply
#88
gmn 17
Time to retire my 2080 TI only 14TF and 11GB VRam. I want the 7800XT custom asus
Posted on Reply
#89
ValenOne
Mr. PerfectHow is the 7800XT using 263W? The 6800XT has 4608 SP and uses 300W, meaning the 7800XT has 16% fewer SP and is only using 12% less power. That's even after a shrink from the 7nm node to 5nm.

The price is better then I was expecting though.
TSMC's 7nm to 5 nm is not a 4 nm or 3 nm node jump. 5 nm is half gen jump. AMD is using TSMC's 4 nm process node for mobile Zen 4 APUs.

For NVIDIA, there's a larger process node jump between Samsung's 8 nm to TSMC's 5 nm with NVIDIA's ADA 4N improved at 5 nm.
Posted on Reply
#90
apoklyps3
This thread should be locked too.
I'm drowning in nvidia fanboys' mouth foam and distorted truth
Posted on Reply
#91
ARF
apoklyps3This thread should be locked too.
I'm drowning in nvidia fanboys' mouth foam and distorted truth
Sometimes the truth is painful for acceptance.
Several more hours you will see it.
Posted on Reply
#92
Beginner Macro Device
Mr. PerfectHow is the 7800XT using 263W? The 6800XT has 4608 SP and uses 300W, meaning the 7800XT has 16% fewer SP and is only using 12% less power. That's even after a shrink from the 7nm node to 5nm.
RDNA3 has a very cursed Perf/W curve and AMD used one of two worst extremums ever on ALL their RDNA3 cards, 7900 GRE possibly, yet unlikely, excluded because I lack data on this one.

The worst extremums are to the left of ~60% TDP and to the right of ~70% TDP. RDNA2 has a stupid flat curve which makes them perfect undervolters. RDNA3 has a steep curve with high floor and low ceiling. Do you mind losing 8% performance? I guess you do. Do you mind also cutting a third of TDP? I don't think you do.



The reason is probably very simple. I assume they wanted to squeeze as much juice from them as they humanly could disregarding the power efficiency aspect because RDNA3 failed extremely hard to deliver proper performance on intended clocks. This led RDNA3 to be a little bit less pathetic compared to Ada and RDNA2 in terms of performance, yet making RDNA3 a very questionable series to say the least.

That said, I expect 7800 XT to run about 500 MHz higher than 6800 XT, and probably about 300 MHz uplift compared to 6700 XT will be seen in 7700 XT. And yes, +10—20% clock almost always means way more than +20% wattage.
Posted on Reply
#93
kapone32
TaishoHere you go:
- AMD is not able to create a product that can compete in raster performance with 4090.
- Claiming "Efficiency leadership" when they already know how efficient Ada is, a couple of weeks ahead of its sales, is not a manipulation but a defiant lie.
- Marketing low manufacturing costs of multi-chip modules, and then launching their GPUs with prices resembling crypto boom launch prices.
- The release is plagued with typical AMD problems like high multi-monitor power draw that based on forum posts is still not fully fixed.
- Vsync power consumption until the monolithic 7600 is almost 2 times higher vs. Ada and worse than Ampere.
- AMD manages to mess up the video playback power draw in 7600.
- Calling 100+ Celsius junction temps on max fan RPM for their reference models working as intended until the huge outrage from the community.
- The 7900 XT is just an upsell product, so basically their lineup is only 2 GPUs: 7900 XTX, which at this price point makes no sense versus RTX 4090, and RX 7600 priced too close to the faster RTX 4060, with far worse value than Navi 2 options.

The only problems with NVIDIA GPUs this gen are prices and carrying on with the plan to make VRAM a luxury good. AMD was quite happy to follow the suit though, locking >8GB RAM behind a solid paywall.
All of what you just said is opinion or issues that have been fixed/resolved. Where I live the 7900Xt is $400 less than the XTX so there is a difference right there. I could also build an entire system of what it costs between a XTX and 4090. It also blows away the 6800XT at 4K high refresh rate Gaming. I also love the whataboutism of VRAM when Nvidia sells a 12GB card for $100 less than the 7900XT (4070TI). So yes I will take the 8GB of VRAM and enjoy my 4K without needing to turn anything on. I expect the 7800XT to be about 10% faster than the 6950XT and as that was 15% faster than the 6800XT we should see about 25% improvement vs a 6800XT.
Posted on Reply
#94
Beginner Macro Device
kapone32I expect the 7800XT to be about 10% faster than the 6950XT
Only if you somehow some way manage to crank 4 GHz out of this 6800 non-XT remake. There is no way it can outperform 6900 series.
Posted on Reply
#95
kapone32
Beginner Micro DeviceOnly if you somehow some way manage to crank 4 GHz out of this 6800 non-XT remake. There is no way it can match 6900 series.
This is 6800XT Transistors 26,800 million from TPU

This is 7800XT Transistors unknown from TPU

Until we know everything both of us could be right and wrong.
Posted on Reply
#97
ARF
Beginner Micro Device6950 XT is obviously faster than 6900 XT and thus, it's faster than 7800 XT. You will need to OC it and, most probably, this won't be enough to outperform 6950 XT.
The RX 6950 XT is 20% faster than RTX 4070.
RX 7800 XT will never be as fast as that, especially that RX 6950 XT should be overclocked, as well, for oranges-to-oranges comparison.

Posted on Reply
#99
ARF
kapone32It all remians to be seen.
How do you think - will it be called epic failure or just a failure?

Posted on Reply
#100
Beginner Macro Device
ARFHow do you think - will it be called epic failure or just a failure?
Don't insult failures!

It's just AMD being AMD.

Posted on Reply
Add your own comment
Aug 14th, 2024 12:26 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts