• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Upcoming Hardware Launches 2025 (Updated May 2025)

purplekaycee said:
what performance difference would the RTX 3090 have over the RTX 3080 TI?
Click to expand...
3080 is like 10% slower than 3090. So a 3080 will be less than 10% slower. Basically indistinguishable outside of benchmarks. Still, 3090's 25GB VRAM can make a difference in compute loads. If you're into that kind of thing.
 
Updated AMD Zen 3 Milan EPYC Server
Updated AMD Zen 4
Updated Intel Alder Lake
Updated Intel Ice Lake Server
Updated Intel Sapphire Rapids
Updated NVIDIA RTX 3060
Updated NVIDIA RTX 3080 Ti
Updated AMD RX 6700 Series
Added AMD Navi 31
Updated Intel Ponte Vecchio
Updated Intel 7 nanometer process
Removed launched RTX 3000 Mobile
 
What we have now is a very weak 3080Ti or a very strong 3080.

2080Ti was +48% 2080, 4352/2944, this now is 10240/8704 like +18% FP32, real 9% in 4K.

4070 should be 20-25% faster than 3080. as early as March 2022 or Nov.

My dream for 12288 Cuda /128 rop, on 7nm, 104 chip, 28B transistors / 65,6 density fits just fine in under 429mm2, 102 be 50% bigger.
 
ppn said:
What we have now is a very weak 3080Ti or a very strong 3080.

2080Ti was +48% 2080, 4352/2944, this now is 10240/8704 like +18% FP32, real 9% in 4K.

4070 should be 20-25% faster than 3080. as early as March 2022 or Nov.

My dream for 12288 Cuda /128 rop, on 7nm, 104 chip, 28B transistors / 65,6 density fits just fine in under 429mm2, 102 be 50% bigger.
Click to expand...
The 3080 is already like 10% behind the 3090, there's not much room for 3080 Ti. 3080 Ti only exists for those that feel uncomfortable with 10GB VRAM.
 
Last edited:
Updated AMD Zen 3 Server / Milan
Updated AMD Zen 4
Updated Intel Rocket Lake
Updated Intel Alder Lake
Added Intel Lunar Lake
Updated RTX 3080 Ti
Updated AMD Radeon RX 6700 & RX 6700 XT
Added AMD Radeon RX 6000 Series Mobile
Added AMD Mining GPUs
Added Radeon Instinct MI200
Updated Intel DG2
Updated Intel Ponte Vecchio
Updated DDR5 Memory
Updated TSMC 3 nanometer
Updated Toshiba 162-Layer NAND Flash
Removed launched products: AMD Cézanne Zen 3 APUs / Ryzen 5000, NVIDIA RTX 3060
 
Why does TPU say RocketLake was based on a GoldenCove Backport?
RocketLake are buyable and clearly show only 512kb L2-Cache, where WillowCove and newer (GoldenCove) have 1.25MB L2 Cache per Core.
 
_Flare said:
Why does TPU say RocketLake was based on a GoldenCove Backport?
RocketLake are buyable and clearly show only 512kb L2-Cache, where WillowCove and newer (GoldenCove) have 1.25MB L2 Cache per Core.
Click to expand...
Because it's possible to backport a core (architecture) while changing the cache size?
In fact, it would be really hard to build something on a bigger process node without cutting back on transistor count.
 
Upcoming hardware paper launches!
 
Is adds and updates for this month? Last update of article was before more of one month.
 
TumbleGeorge said:
Is adds and updates for this month? Last update of article was before more of one month.
Click to expand...
p1kalmig2k671.jpg


just need to find a bit of time :)
 
* "Article". :) But don't spend time on fixing that versus the list!
 
Updated

Added AMD Ryzen 5000G APUs for Desktop (Cezanne)
Updated Intel Sapphire Rapids
Updated Zen 3 Threadripper 5000
Updated Intel Meteor Lake
Updated Radeon RX 6800 Mobile
Updated NVIDIA RTX 3050 / 3050 Ti
Updated NVIDIA RTX 3070 Ti
Updated NVIDIA RTX 3080 Ti
Updated Intel Ponte Vecchio
Added AMD X570S Chipset
Updated DDR5 System Memory
Updated TSMC 4 nanometer
Removed launched products: AMD Zen 3 Server Milan, Intel Ice Lake Server, Intel Rocket Lake, RX 6700 XT, Intel 500 Series chipsets
Removed cancelled products: Intel Core i9-10990XE, Radeon RX 5000 variants
 
Looks like we just gotten over an exiting or interesting March with the 11th Gen Intel CPU releases and new Z590 mobos. The next big 'tech-entertainment' will be just in the months ahead with Alder Lake seeing the light of day. Another reviewer delight, reporting and testing period. (PCIe 5.0, DDR5 and the i9 is tauted to have 16 cores?) Then all the motherboard manufacturers breaking out as well (for their second time this year) with new LGA 1700 boards. Then topping all of this will be the AIO, memory, storage and PSU manufactures trying to proffer upgraded and or matching 12th Gen products. So many new products coming soon, so many new tests, so many new reviews, so many new opinions. As such the remainder of 2021 should confirm that we are living in a very interesting time. If only now the GPU fiasco would solve itself later this year that would be the icing on the cake.
 
Some of the sources don't seem to understand how memory buses work, and I'm surprised that no-one seems to have picked up on this. Some people probably have, but I've not seen any mention of it so far.

  • 12 GB GDDR6
  • 192-bit or 256-bit memory bus
Click to expand...
If it's 12GB, it's 192 bit.

  • 58 SM, 7424 cores, 320-bit memory, 16 GB VRAM, GDDR6 or GDDR6X
Click to expand...
If it's a 320-bit bus, it's not 16GB. It would be 10GB.
If it's 16GB, it is 256-bit.

  • Most likely Navi 12 without HBM2
Click to expand...
You can't do that. Navi 12's memory controller and PHY are designed for HBM2, and are not compatible with GDDR6. Cryptomining is very bandwidth-sensitive, and profitability is heavily dependent on minimising power usage, so AMD would want to use HBM2 even if they had the option not to - it would be worth it, especially with GPU prices as inflated as they are at the moment. Navi 10 effectively is "Navi 12 without HBM" (i.e. 40 RDNA1 CUs with GDDR6 instead of HBM2). If it's Navi 10, call it Navi 10. If it's Navi 12, it will use HBM2.

The bandwidth mismatch that results when you use memory chips of different sizes means Nvidia is not going to use a 4x1GB+4x2GB layout, which is what would be needed to fit 12GB onto a 256-bit bus. There was a huge backlash against Nvidia when they last tried something like this (with the GTX 970).
 
Gradius2 said:
There is nothing about Ryzen 6000 ?
Click to expand...
Ryzen 6000 should be using Zen 4, so see that for info.
Although I've heard rumors that AMD are doing a Zen 3+ and that 7000 is the first to use Zen 4? I haven't been paying that much attention to this stuff lately so I'm not 100% sure.
 
Added AMD Zen 3+ / Rembrandt APUs
Updated AMD Zen 5
Updated Intel Alder Lake
Updated Intel Jasper Lake
Updated Intel Sapphire Rapids
Updated NVIDIA RTX 3070 Ti
Updated NVIDIA RTX 3080 Ti
Updated AMD RDNA3
Added AMD Radeon Pro W6800
Updated Intel DG2
Updated Intel Ponte Vecchio
 
Updated AMD Cezanne Ryzen 5000G
Updated AMD Zen 3+
Updated AMD Zen 4
Updated AMD Zen 5
Updated Intel Alder Lake
Added Intel Raptor Lake
Updated Intel Meteor Lake
Updated Intel Sapphire Rapids
Added AMD Radeon RX 6600 / RX 6600 XT
Added AMD Radeon RX 6400 Series
Updated AMD Radeon Instinct MI200
Updated AMD RDNA3
Updated Intel DG2
Updated HBM3 Memory
Updated TSMC 4 nanometer
Added TSMC 1-nanometer
Updated Samsung 5 nanometer
Updated Micron 176-layer NAND
Updated Toshiba/WD PLC NAND Flash
Updated Samsung 176-layer NAND
Removed launched products: RTX 3070 Ti and RTX 3080 Ti, RX 6700 XT, RX 6000 Mobile, Radeon Pro W6800, X570S chipset, Micron 128-layer NAND flash
 

​

AMD Radeon RX 6600 / RX 6600 XT [added]​

  • Release date: 2021
  • Based on 7 nanometer Navi 23 GPU
  • 11 billion transistors, 237 mm² die size
  • 32 CUs, 2048 cores, 128 TMUs, 32 ROPs
  • RX 6600 non-XT: 1792 cores, 112 TMUs, 32 ROPs
  • 8 GB 128-bit GDDR6 + 32 MB on-die L3 cache
  • Interface PCIe 4.0 x8 or x16
  • Sources

Then

AMD Radeon RX 6500​

  • Release date: unknown
  • 40 compute units / 2560 stream processors
  • 192-bit GDDR6 memory
  • 7 nanometer production process
  • RDNA2 architecture
  • Codename "Navy Flounder"
  • Below $250
  • Sources

I think that needs some updating.............maybe 22CUs / 1408 shaders
 
mechtech said:
I think that needs some updating.............maybe 22CUs / 1408 shaders
Click to expand...
Possibly, yeah. Check the source links, I'm going by those
 
40 CU chip is already used in RX 6700 XT, it can't be RX 6500 ...

Navy Flounder => Navi 22 => RX 6700 XT
Dimgrey Cavefish ? Navi 23 ? RX 6600/6500 XT
Beige Gooby ? Navi 24 ? RX 6500/6300 XT
 
Updated Intel Alder Lake
Updated Intel Sapphire Rapids
Added Intel Sapphire Rapids HEDT
Updated AMD Radeon RX 6600 / RX 6600 XT
Updated AMD Radeon Instinct MI200
Updated Intel DG2 Discrete GPU
Updated Intel Ponte Vecchio
Added Intel Z690 Chipset
Added Intel B660 & H610 Chipsets
Added Intel W790 Chipsets
Updated TSMC 3 nanometer tech
Updated Samsung 5 nanometer tech
Updated Samsung 3 nanometer tech
Updated PCI Express 5.0
 
You must log in or register to reply here.
Back
Top