The very first details about AMD's new mainstream Radeon RX 7600 graphics card first emerged last week when several stores in Asia began to offer the product both in retail and on the marketplace. VideoCardz today published a GPU-Z screenshot that reaffirms the general specifications of the device and uncovers some additional bits of info. As with all leaks, take the information with a grain of salt.
AMD's Radeon RX 7600 is based on the company's Navi 32 XL graphics processor with 32 compute units or 2048 stream processors mated with 128 texture units and 32 raster operating units that can burst from the default 2250 MHz game clock all the way to 2655 MHz, at least according to the GPU-z screenshot. The GPU has a 128-bit GDDR6 memory interface that offers a bandwidth of 288GB/s as well as a PCIe 4.0 x8 interface to connect to the host. As for the card, it carries 8GB of memory.
In real-world scenarios, AMD's Radeon RX 7600 is said to be able to overclock itself to 2.85 GHz, which translates into 23.3 FP32 TFLOPS, which is in line with compute throughput of Radeon RX 6900 XT, one of the best graphics cards around. Meanwhile, compute throughput of AMD's RDNA 3 architecture-based GPUs does not directly transform into real-world gaming performance, so we do not expect the Radeon RX 7600 to beat the previous-generation flagship offering.
VideoCardz stresses that the GPU-z software currently does not officially support the retail Radeon RX 7600 graphics cards. Meanwhile, it does support and can recognize AMD's Navi 33 GPU as it has to correctly recognize Radeon RX 7700/7600 mobile series. Thus, while the software might fail to identify the card, it still manages to display the accurate GPU specifications, the website concludes.
We would still add that since the information comes from unofficial channels, one should take it with precaution. Yet, keeping in mind that AMD and its partners are rumored to start selling their Radeon RX 7600 products next week, the probability that the current leaks are wildly inaccurate is extremely low.
Slight typo error. It's NAVI 33 XL GPU core die, not 32.
So true. Intel already dropped the price of the A750 to $229 just few months ago, and I think it would be hard for Intel to compete with AMD and Nvidia now, even in the mainstream/mid-range GPU market segment.
Some custom A750 were selling for 229 USD on Newegg, so if Intel drops it to 200 USD, then I'm not sure how will they manage to keep with the GPU market price trend, or the sales.
I mean Intel is a third NEW entrant in the GPU market business, discrete that is, so I think that's the price one has to pay for being a fresher in the discrete gaming GPU market. But since Intel is the third contender so maybe they will take some more time to catch up with both AMD and Nvidia.
I'm assuming the next-gen Battlemage "XE2" gaming GPUs might give Intel a strong foothold in the discrete gaming GPU business/market share, but that remains to be seen. Alchemist actually failed to capture the market fully.
You got a defective A750 GPU, or only the box was open ?
For other low-end/mainstream cards from AMD and Nvidia I can easily upgrade though. But I doubt the ARC GPU will properly work on my old HASWELL platform.
I updated my BIOS, updated drivers, tried shuffling a few things in BIOS, still no improvement.
I put my GTX1050 back in with no other changes, returned the A750, just passed 100h of uptime, so I can be relatively certain that nothing I changed in BIOS while attempting to 'fix' stability issues with the A750 affected my system's stability with a known-good GPU installed.
Without a second system with reBAR to confirm that instability follows the hardware, the only conclusion I can come to is that the A750 I got was probably defective. If the A750 officially drops to $200, I may buy another and get to confirm my original suspicion based on whether the crashes return with a different presumably good unit. If the crashes do return, then my next troubleshooting step may be a new SSD for a fresh Windows 11 install to eliminate software and driver leftovers as a variable.
RX7900XTX: 61.42 TFLOPS
RX6900XT: 23.04 TFLOPS
RX6600XT: 10.60 TFLOPS
RX7600: 23.3 FP32 (best case, it seems)
Ratios: 6900XT/6600XT is 2.17 times and it's about 57% of its performance. The ratio on the RX7900XTX and RX7600 is 2.63 times, so it's going to be about 47% of its performance or whereabouts? That would put the card in between 6700XT and 6750XT territory. Best case 6750XT, but probably more like a 6700XT or around/under it.
Links I used for this napkin math:
- I have a 6600XT and have found it to be very capable at 1440p
That's kind of odd, since I have not heard any of the new Intel GPUs dying this early, or are defective/DOA. Could it be your PSU causing the crashes ?
My PC runs fine for months on end with the most aggressive CPU C-states enabled (spending ~80% of its time in C7 right now) with my GTX1050, which makes me quite confident my PSU can handle harsh ~100W transients perfectly fine all day, every day.
My PSU is a 650W SilverStone Strider Platinum which I received as a gift years ago. Not the most praised PSU brand but still far more high-end than anything I would have bought for myself and far more high-end than what most people buying an A750 are likely to have.
Since it was an open-box, my guess is whoever bought it first ran into similar issues, didn't have the patience to try to sort it out, returned it for whatever reason, store gave it a quick whirl to check that it was still working, put it back on the shelf hoping the next guy (me) would either have better luck or come back with a more detailed account of problems encountered with the card.
Maybe you haven't heard about A-series GPUs dying because there aren't that many out there in the first place. Early failure rates for consumer electronics are usually in the 1-2% range, failed A7xx should definitely be out there even if you haven't seen any yourself.