Radeon RX 7600 final specifications leaked, 165W TDP & 2625 MHz boost
Unofficial sources reveal key details about the AMD Radeon RX 7600 graphics card. This latest offering, founded on the RDNA3 architecture, suggests higher power needs than its predecessor.
The RX 7600 boasts a Typical Board Power (TBP) of 165W, necessitating a minimum 550W power supply - a 33W increase from the RX 6600 GPU, which is built on the RDNA2 framework. AMD has included a comparative power consumption chart in their reviewer's guide to illustrate this difference.
Several other details have emerged shortly before the product's launch, confirming the complete specifications including clock speeds and memory configuration. Though not entirely new - as AMD has already released similar data for its mobile series - these specifications are specific to the desktop version. The card is expected to have a game clock of 2250 MHz and a boost clock of 2625 MHz. These specifications are in line with previous leaks, although the boost clock of the reference model is reported to be slightly lower. The card will incorporate a full configuration of 32 Compute Units and 2048 Stream Processors for the Navi 33 GPU.
The leaked information also indicates that the RX 7600 will feature quicker memory than the RX 6600. It will house 18 Gbps GDDR6 modules, offering a maximum bandwidth of 288 GB/s. AMD's data suggests an effective bandwidth of 477 GB/s when taking into account the Infinity Cache, a modest improvement over the RX 6600's 413 GB/s. Despite the increase in memory speed, the memory bus remains at 128 bits. Like its predecessor, the RDNA3 model will maintain an 8-lane limit on its interface.
The Radeon RX 7600's official launch is anticipated on May 25th. The RX 7600's core GPU, while based on the advanced RDNA3 graphics architecture, continues to utilize the 6 nm (DUV) silicon fabrication process from the prior "Navi 24" GPU. The silicon transistor count is approximately 13.3 billion, around 2 billion more than the 7 nm "Navi 23" in the RX 6600 series, while the die-size is constant at 204 mm². The GPU maintains a PCI-Express 4.0 x8 host interface and a 128-bit GDDR6 memory interface.
The RX 7600 incorporates 32 RDNA3 compute units (CU), translating to 2,048 stream processors. It shares the dual-issue instruction rate characteristic with the RX 7900 series. Additionally, it boasts 32 Ray Accelerators and 64 AI Accelerators. The GPU has 128 TMUs and 64 ROPs, in addition to 32 MB of second-generation Infinity Cache memory. The 8 GB of GDDR6 memory operates at 18 Gbps, enabling a memory bandwidth of 288 GB/s across the 128-bit memory bus. According to AMD, the effective bandwidth is further enhanced when combined with the on-die cache.
Radeon RX 7000 Specifications | ||||
---|---|---|---|---|
Radeon RX 7600 | RX 6650 XT | RX 6600 XT | RX 6600 | |
GPU | Navi 33 XL | Navi 23 KXT | Navi 23 XT | Navi 23 XL |
GPU Clusters | 32 CUs | 32 CUs | 32 CUs | 28 CUs |
FP32 Cores | 2048 | 2048 | 2048 | 1792 |
Base Clock | 1720 MHz | 2055 MHz | 1968 MHz | 1626 MHz |
Game Clock | 2250 MHz | 2410 MHz | 2359 MHz | 2044 MHz |
Boost Clock | 2625 MHz | 2635 MHz | 2589 MHz | 2491 MHz |
Memory | 8GB G6 | 8GB G6 | 8GB G6 | 8GB G6 |
Memory Bus | 128-bit | 128-bit | 128-bit | 128-bit |
Memory Speed | 18 Gbps | 17.5 Gbps | 16.0 Gbps | 14.0 Gbps |
Bandwidth | 288 GB/s | 280 GB/s | 256 GB/s | 224 GB/s |
Board Power | 165W | 180W | 160W | 132W |
Power Connectors | 1x 8-pin | 1x 8-pin | 1x 8-pin | 1x 8-pin |
PCIe Interface | Gen4 x8 | Gen4 x8 | Gen4 x8 | Gen4 x8 |
MSRP | $399 | $379 | $329 | |
Launch Date | May 25, 2023 | May 10, 2022 | August 11, 2021 | October 13, 2021 |
Custom Radeon RX 7600 Graphics Cards Revealed Ahead of Launch - 05/22/2023 08:48 AM
VideoCardz unveiled images and details of custom design Radeon RX 7600 graphics cards. ASRock and ASUS are among the manufacturers with factory overclocked models, providing a sneak peek just days b...
AMD Radeon RX 7600 reference card pictured - 05/22/2023 08:44 AM
AMD's latest offering, the RX 7600 rerference, is a compact graphics card distinguished by its dual fan design and modest size, measuring less than 21 cm. ...
Leaked Details Emerge for AMD Radeon RX 7600 GPU - 2048 Stream Processors - 05/19/2023 08:37 AM
As the late May launch date approaches, leaks regarding the AMD Radeon RX 7600 GPU continue to surface. This entry-level model is expected to compete with NVIDIA's upcoming GeForce RTX 4060 graphics ...
Modder Pushes Radeon RX 7900 XTX Overclocking to New Heights: Achieving 3467MHz Speeds - 05/17/2023 08:07 AM
A modder, identified as jedi95, has recently achieved a significant accomplishment by overclocking the AMD Radeon RX 7900 XTX graphics card. By ingeniously modifying the card, jedi95 was able to raise...
Custom Radeon RX 7600 GPUs Listed by Canadian Retailer: Details and Prices - 05/16/2023 06:33 PM
Canadian retailer, PC Canada, has recently listed custom Radeon RX 7600 graphics cards from popular brands Sapphire and MSI, and we've got all the details for you....
Senior Member
Posts: 8254
Joined: 2020-08-03
it's not fake, but not entirely true either. it's true for certain resolutions only, as this relies on cache hit/miss rate, which depends on resolution.
it's only interesting from technical point of view, as 32mb l2 appears to be 16% more effective than 32mb l3. (554 vs 476 effective measured with 18gbps 128-bit at 1080p)
Senior Member
Posts: 22345
Joined: 2008-08-28
it's not fake, but not entirely true either. it's true for certain resolutions only, as this relies on cache hit/miss rate, which depends on resolution.
it's only interesting from technical point of view, as 32mb l2 appears to be 16% more effective than 32mb l3. (554 vs 476 effective measured with 18gbps 128-bit at 1080p)
So it just effects the target resolution? As soon as you up to 1440p performance hit is huge. It would be interesting if the cache is bigger would the performance hit at higher resolutions would be less severe.
Senior Member
Posts: 8254
Joined: 2020-08-03
yup. read nvidia's paper on l2 cache and effective bandwidth. they quote it's 554gb/s effective, because of 50% cache hit rate.
that 50% hit rate is 1080p tho.
also, this 7600 is 6nm, more room for milking refreshes I guess.
Senior Member
Posts: 8254
Joined: 2020-08-03
yes, it depends on cache hit/miss ratio. 72mb on 4090 prolly has the same 50% as 32mb at 1080p, they must have aimed for that 50% on each sku cause the numbers are weird.
it also depends on cache latency and bandwidth. l2 will have lower latency than l3. dunno about bandwidth they're both using.
Senior Member
Posts: 22345
Joined: 2008-08-28
Still no price? Both AMD and Nvidia should stop that effective memory bandwidth crap. Add some cache fine but rx7600 does not have 476gbps, its fake.