Graphics Cards - Page 33

Last year, we got our first look at Thermal Grizzly's WireView Pro tool for GPUs using the 12VHPWR and 12V-2x6 power connectors. A device for high-end GeForce RTX graphics cards, this tool offers valuable data by measuring and monitoring temperature and power directly at the connection point.

Fast-forward to Computex 2025, and we've seen more cases of melting cables for the 575W GeForce RTX 5090 emerge. Investigations, like this one from De8auer, have discovered a potential imbalance in the GPU's power getting evenly spread across all of the individual cables found on the newer 12V-2x6 power connector.

With that, Thermal Grizzly was back at Computex with a prototype of the new WireView Pro II, which has been designed to monitor and detect issues related to the new power connector. And yes, this includes the ability to measure per-pin current to detect imbalance loading and warn users in advance of any potential failure due to power issues.

ASUS had an awesome booth at Computex 2025 this year, with one of (the many) new product unveilings being the new ROG XG Station 3, a new Thunderbolt 5-powered external GPU solution.

The new ASUS ROG XG Station 3 (with Thunderbolt 5) is an external graphics dock that boosts gaming and performance on your laptop, and it's not just for gamers: content creators, and higher-end AI workloads require much faster GPUs than you can get in a laptop... as long as you've got Thunderbolt 5, you can enjoy up to 80Gbps of bandwidth.

The external GPU solution is compatible with NVIDIA's new GeForce RTX 50 series "Blackwell" GPUs as well as AMD's new fleet of Radeon RX 9000 series "RDNA 4" graphics cards.

Priced at over $10,000 in some markets, the new RTX PRO 6000 GPU is a new workstation GPU that can be viewed (in some ways) as a GeForce RTX 5090 with 96GB of VRAM. This thing is a monster, and we got to go hands-on with it at PNY's booth at Computex 2025. Well, hold it and take a look at its physical design.

Yes, it looks almost identical to the GeForce RTX 5090 Founders Edition (check out our review here), but it has an all-black finish and an 'RTX PRO 6000' label. The cooler design and form factor are identical, with a two-slot thickness, custom PCB, and NVIDIA's new Double Flow Through cooling.

The big thing here, of course, is that it packs a whopping 96GB of fast GDDR7 memory and the full GB202 GPU chip with 24064 CUDA Cores, compared to the RTX 5090's 21760 CUDA Cores. To fit all of that memory inside the card, NVIDIA has placed 48GB on either side of the PCB.

INNO3D is a brand we've grown to admire over the years because it has consistently created compact and powerful GeForce RTX graphics cards. A two-slot RTX 5080 is one recent example; another is a dual-fan RTX 5070 that could fit inside the tiniest PC gaming case. If you're after something small, compact, and with a clean, minimal design, then INNO3D X2 or X3 cards are worth looking into.

At Computex 2025, INNO3D showed us its INNO3D GEFORCE RTX 5090 X3, in white, inside a compact ITX case. As a GeForce RTX 5090, it's still fairly big, as expected, but it was cool to see it fit inside an ITX case, keeping cool while playing DOOM: The Dark Ages at over 300+ FPS in 4K with max settings.

This is INNO3D's first White-colored GeForce RTX 5090, which the company notes is quickly becoming Asia's most popular color option for PC gaming components. We've lost count of the sheer number of white-colored PC gear we've seen this week at Computex 2025, so that remark makes a lot of sense.

MSI introduced its new Expert series GPU design last year with the MSI GeForce RTX 4080 SUPER EXPERT (check out our full review here). Outside of the striking RGB-free all-metal shroud, what made the Expert stand out from the pack, including MSI's other cards, was that it incorporated a push-and-pull cooling design modeled after NVIDIA's Founders Edition cards for the RTX 30 and RTX 40 Series.

The push-and-pull design puts a large fan on either side of the GPU to increase airflow and cooling in a traditional horizontal orientation. And when you pair that with the MSI Expert brushed metal finish and metallic vents, you've got a GPU that still looks unlike anything else.

Fans of the Expert design will be pleased to know that it's set to return with the new GeForce RTX 5080 16G EXPERT OC, which was announced and showcased at Computex 2025.

At Computex 2025, MSI gave us a quick tour of its GPU Vision display, which showcased a selection of graphics card and cooler designs that its team is developing. From the new flagship RTX 5090 SUPRIM TITANIUM EDITION, which features a gold-colored exterior made from lightweight titanium, to the return of the classic open-air Cyclone design, now for the GeForce RTX 5070.

MSI also showed us another blast from the distant GPU past, with a new RTX 5060 Ti 16GB TWIN FROZR 2025 that captures the look of this classic GeForce GPU. As we're talking about more fabulous designs from an era when graphics cards were nowhere near as powerful, complex, and power-hungry as they are in 2025, naturally, MSI had to reengineer the thermal designs to accommodate the GeForce RTX 50 Series.

The GeForce RTX 5060 Ti 16G TWIN FROZR 2025 OC includes shifting to a copper-aluminum composite material for the fins, as the traditional all-aluminum solution simply wouldn't work. Also, there's now a nickel-plated copper baseplate for heat transfer and MSI's new STORMFORCE fans. It's retro, sure, but it's also stylish and impressive.

At Computex 2025, we got the chance to visit ASRock behind closed doors to unbox and check out its new flagship mainstream RDNA 4 GPU, the ASRock Radeon RX 9060 XT 16GB Steel Legend. We're familiar with this design at TweakTown, having recently reviewed the impressive ASRock Radeon RX 9070 Steel Legend OC.

Here, we get to unbox a model straight from the production line. It is ASRock's flagship OC model for AMD's new Radeon RX 9060 XT, which was announced a day ago and will hit shelves around the world on June 5. It ships with a Boost Clock speed of 3320 MHz, a sizable increase over the 3130 MHz of the reference design. This should make it one of the fastest Radeon RX 9060 XT cards on the market.

As the more mainstream version of the 9070 Steel Legend, some key differences are that it's lighter and the bracket only features three display ports. This is standard for the Radeon RX 9060 XT, which all feature 2 x DisplayPort 2.1a and 1 x HDMI 2.1b ports. According to AMD, the reason for this is that only a fraction of people actually use more than three displays.

AMD has formally announced the Radeon RX 9070 XT at Computex 2025. The mainstream RDNA 4 GPU will compete with NVIDIA's GeForce RTX 5060 Series. Like the flagship Radeon RX 9070 XT released earlier this year, the mainstream RDNA 4 GPU brings enhanced ray-tracing performance and the new AI-powered FSR 4 to mainstream Radeon gaming.

The Radeon RX 9060 XT will be available in two versions: one with 8GB of GDDR6 memory and one with 16GB of GDDR6 memory, both on a 128-bit memory interface. Outside of memory, the specs are identical with 32 Compute Units or CUs and a reference Boost Clock speed of 3.13 GHz. Pricing will start from $299 and $349 USD, respectively. During AMD's big reveal, the company noted that the Radeon RX 9060 XT 16GB will be the most powerful gaming GPU for under $350, which is 6% faster than the GeForce RTX 5060 Ti 8GB on average.

At Computex 2025, we got to check out Sapphire's full range of Radeon RX 9060 XT 8GB and 16GB offerings, covering reference Pulse designs, white-colored Pure models, and the flagship triple-fan Nitro+ model. Interestingly, all will be available in OC form. Let's take a look.

AMD has fully revealed the new Radeon RX 9060 XT at Computex 2025, confirming the details we've been waiting for - price and availability. As already confirmed, the mainstream RDNA 4 GPU is set to become available in two models, one with 8GB of GDDR6 VRAM and one with 16GB, on a 128-bit Memory Interface.

Outside of the VRAM capacity, both versions ship with the same 32 Compute Unit chip (half the size of the flagship Radeon RX 9070 XT) that includes 32MB of AMD Infinity Cache and a Boost Clock speed of up to 3.13 GHz. The Radeon RX 9060 XT 16GB features a higher TBP (Total Board Power) rating of 160W compared to the Radeon RX 9060 XT 8GB's 150W.

As for the pricing and release date, both cards are set to launch on June 5. The suggested retail price of the Radeon RX 9060 XT is $299, and the Radeon RX 9060 XT 16GB is $349 USD. AMD notes that models from Acer, ASRock, ASUS, GIGABYTE, PowerColor, Sapphire, Vastarmor, XFX, and Yeston are on the way.

NVIDIA's beefed-up GeForce RTX 5080 SUPER graphics card is back in the headlines, with leaked specs suggesting we can expect 24GB of GDDR7 memory clocked at 32Gbps, 10752 CUDA cores, and 400W+ of power.

It was only earlier this month that leaked Moore's Law is Dead said that we were to expect the new RTX 5080 SUPER with 24GB and the new RTX 5070 SUPER with 18GB later this year, with between 7% and 15% performance gains over their non-SUPER counterparts.

But in a new post by inside @kopite7kimi on X, we've got some fresh specs that include the PG147-SKU35, GB203-450-A1 GPU, 10752 CUDA cores, 24GB of GDDR7 clocked at 32Gbps on a 256-bit memory bus, and a rather hefty 400W+ of power consumption.

The new AMD Radeon AI PRO R9700 GPU has been announced at Computex 2025, bringing RDNA 4 to AI-powered workstations. With 32GB of GDDR6 memory, 96 TFLOPS of Peak Half-Precision, and 1531 TOPS of INT4 Sparse AI performance, you're looking at up to 2X better performance than the previous generation's AMD Radeon PRO W7800 32GB GPU.

With 32GB of VRAM, AMD notes that the Radeon AI PRO R9700 is better equipped for running advanced local text-to-image AI models and LLMs like DeepSeek R1 Distill Qwen 32B Q6 and Mistral Small 3.1 24B Instruct 2503 Q8. With expanded AMD ROCm on Radeon, there'll be support for a "broader range of AI and compute workloads."

The first chart in the presentation for the new AI workstation GPU compares the average tokens per second performance against the GeForce RTX 5080, showing an increase of up to 496%. Check out the chart below.

Ahead of Computex 2025 kicking off, AMD held a pre-brief with the media, during which it formally announced and confirmed the specs for its mainstream RDNA 4 GPU, the Radeon RX 9060 XT. Although we'll still have to wait for official benchmark performance, pricing, and retail availability (as far as we can tell, the cards are all good to go), here's everything we know so far - straight from the source.

First, we have some general performance info - potentially great news. As we've seen with the flagship Radeon RX 9070 XT, RDNA 4 has been a game changer for ray-tracing performance on an AMD GPU, which will be the case with the Radeon RX 9060 XT. The Radeon RX 9060 XT's 32 RDNA 4 Compute Units or CUs "double ray-tracing throughput" compared to the Radeon RX 7600.

With 821 TOPS of AI performance (INT4 Sparse), the Radeon RX 9060 XT will be able to leverage AMD's AI-powered FSR 4, finally an alternative to NVIDIA DLSS, to deliver more 1080p and 1440p gaming performance while minimizing the impact on image fidelity.

NVIDIA's mainstream GeForce RTX 5060 GPU is out this week, and at Computex 2025, we've already seen countless models and designs from the company's partners, many of which look familiar. However, one particular RTX 5060 caught our eye and was found at ZOTAC's booth.

You're looking at a prototype and a card still in development. Alongside its low-profile design, it's also one of the smallest RTX 50 Series GPUs you're likely to find - small enough to fit in the palm of your hand.

It's a reference spec model with three fans, and it looks like a miniature version of ZOTAC GAMING's triple-fan SOLID series of cards for the GeForce RTX 50 Series. The fans are tiny, and it's so light that at a glance, you'd probably mistake it for a toy of a real GPU.

NVIDIA's latest graphics driver has fixed a bug with its RTX 5000 laptop graphics cards which caused them to underperform somewhat.

VideoCardz highlighted the cure for the sluggishness related only to mobile Blackwell GPUs, which meant their TGP wasn't being pushed as high as it should've been, leading to a drop in performance.

Reviews of gaming laptops with RTX 5000 GPUs inside had uncovered this oddity with power delivery, but it wasn't something NVIDIA commented on. Oddly, as VideoCardz pointed out, the official bug fix was only documented via the notes on NVIDIA's GeForce forums, and not with the release notes accompanying the actual driver.

At Computex 2025, ZOTAC showcased a prototype of its new flagship GeForce RTX 50 Series GPU - the ZOTAC GAMING GeForce RTX 5090 ArcticStorm AIO. As the name suggests, this is an RTX 5090 that will ship with an all-in-one liquid cooling solution, with a custom waterblock with a full metal backplate and a large 360mm radiator with three 120mm fans.

With liquid cooling, it will ship with the fastest out-of-the-box overclocked specs for the company's entire range of RTX 5090 GPUs, making it the fastest ZOTAC GAMING GPU ever made. The version we saw up close was a prototype; however, the physical design itself is finalized.

The company is also planning to release its flagship GeForce RTX 5090 AIO card sometime in Q3 2025, though no specific release date or price is available at this time. Naturally, this will be the most expensive GPU in ZOTAC GAMING's line-up, with the company referring to it as the "crown jewel" of its Computex 2025 GPU showcase.

Gamers Nexus and Hardware Unboxed have released new videos outlining NVIDIA's attempted manipulation of coverage for its GeForce RTX GPUs based on their experiences with the company. This comes to a head with this week's release of the GeForce RTX 5060, where media weren't given access to pre-release drivers for the mainstream RTX Blackwell graphics card.

And with the GeForce RTX 5060 launch and drivers becoming available during Computex 2025, when most media cannot benchmark and complete testing, it's not a good look for NVIDIA's approach to the latest RTX 50 Series launch. I can confirm the latter, as I've had a GeForce RTX 5060 in my possession for over a week with no drivers to complete a review for TweakTown ahead of the show.

Gamers Nexus's 'NVIDIA's Dirty Manipulation of Reviews' video highlights their detailed experiences with NVIDIA. It outlines how the company wanted to see coverage of DLSS 4 technologies, like Multi Frame Generation, to gain access to certain engineers or employees for its highly technical, in-depth coverage.

At Computex 2025, Intel announced its new Battlemage Arc Pro GPUs for AI workstations, with multiple models and variants from its partners, such as ASRock, Sapphire, MaxSun, and more. The Arc Pro B-series, powered by the same Battlemage architecture as the desktop Intel Arc B580 gaming GPU, includes two models - the Intel Arc Pro B50 16GB, starting from $299, and the Intel Arc Pro B60 24GB for around $500 per GPU.

That's per GPU because the Arc Pro B60 is designed for systems that can house multiple B60 GPUs in scalable AI inference workstations dubbed 'Project Battlematrix.' We have to see these systems up and running at Computex 2025, where up to 8 Intel Arc Pro GPUs offer up to 192GB of VRAM to support 70B+ parameter models with full software support and optimization.

According to the performance chart below, the new Intel Arc Pro B50 is set to offer an efficient mainstream option for AI workstations compared to the NVIDIA RTX A1000.

The ROG ASTRAL is ASUS's new flagship GeForce RTX GPU, a quad-fan beast designed for the GeForce RTX 5080 and GeForce RTX 5090. Earlier this year, we reviewed the ROG Astral GeForce RTX 5080 and praised its "incredible 4K gaming performance" that delivers up to double-digit gains over NVIDIA's Founders Edition reference design.

With the arrival of DOOM: The Dark Ages on PC and consoles, ASUS has partnered with Bethesda and id Software to create a custom ROG Astral GeForce RTX 5080 DOOM Edition GPU to celebrate the latest release in the iconic series. With custom colors that include a couple of fans looking like the Slayer's, gold fins, and a custom backplate, it's an incredible-looking GPU.

At Computex 2025, we got to go hands-on with one of these rare beasts (the ROG Astral is one of the largest GPUs on the market), as only 666 are being made, and even used it to play some DOOM.

GIGABYTE unveiled its new AORUS RTX 5090 AI Box at Computex 2025, where we were on the ground snapping some photos of the new external GPU solution.

The company might not be the first with a high-end external GPU solution -- something ASUS revealed earlier this year with its ROG XG 2025 external GPU -- but it has the fastest eGPU solution with its new AORUS RTX 5090 AI Box. GIGABYTE's new AORUS RTX 5090 AI Box is an external GPU box with a built-in ATX 3.1 PSU and a super-fast Thunderbolt 5 interface blasting up to 120Gbps of bandwidth thanks to the use of asymmetric bandwidth boost.

This means that the host to the GPU transmit stream speeds along at 120Gbps, while the host to the GPU stream runs at 40Gbps. The new AORUS RTX 5090 AI Box has a GeForce RTX 5090 graphics card inside, and an AIO cooler to keep thermals under control.

The GeForce RTX 5060 is launching today, starting from $299. Although our review won't be ready for a while due to no pre-release drivers being available, NVIDIA is sharing performance data for some of the most popular games running on the mainstream GPU. 100+ FPS running in 1080p with Max Settings, which includes 223 FPS for DOOM: The Dark Ages.

However, with DLSS 4 enabled using the Quality preset for Super Resolution and the maximum Multi Frame Generation setting enabled in most games, which is 4X, three AI frames are generated for each single DLSS 4 Super Resolution frame - you're looking at a big asterisk.

In addition to the above, DOOM: The Dark Ages also uses the Ultra preset (not Nightmare) with a custom texture pool size to account for the GPU's 8GB of VRAM. The GeForce RTX 5060, featuring 8GB of fast GDDR7 memory, means some games need settings lowered to accommodate the VRAM capacity limitation.