Latest news with #graphicscard
Digital Trends
20-07-2025
- Digital Trends
GPUs with 8GB VRAM are 2025's real budget trap
It has become increasingly hard to justify buying a graphics card with only 8GB of VRAM. Once the standard for midrange GPUs, it is now a hard bottleneck in modern games and the cracks are showing more than ever. Whether you're planning to upgrade or building a new budget PC, it's time to reconsider what you actually get with that 8GB label. Let's be clear, this isn't about raw performance alone. As we pointed out during the launch of the RTX 5060, the problem lies less in core count or clock speeds, and more in how 8GB of VRAM caps your long-term usability. Even if the card is decent at 1080p right now, it's the kind of hardware decision you'll regret two years from now. Recommended Videos Take Indiana Jones and the Great Circle, a 2024 release that made headlines for its high system requirements. As per testing done by Computer Base, the game chews through VRAM at 1080p using ultra textures, forcing cards like the RTX 4060 and 4060 Ti 8GB to drop frames or crash entirely. That's not just limited to a particular title, it's becoming the new norm. Games like The Last of Us Part I, Hogwarts Legacy, and Alan Wake 2 are similarly harsh on VRAM, especially at higher settings. And no, simply lowering textures doesn't always 'fix' the problem as it can still result in texture pop-in, long asset loading, and a generally compromised experience. VRAM is the real limiting factor The conversation around GPU performance often fixates on frames per second, but in 2025, VRAM capacity is increasingly the thing separating playable experiences from broken ones. When modern game engines request more than your card can handle, it results in poor performance including stutters and crashes. It doesn't matter if your 8GB GPU technically has enough shader power, it becomes a glorified bottleneck. Worse still, newer cards with 8GB are often misleadingly marketed. Nvidia's RTX 5060 and 5060 Ti (8GB variants) look appealing on paper, with Blackwell efficiency and support for DLSS 4 with multi-frame generation. But when they choke on big textures or fail to keep up in open-world games, the real-world experience often falls short. AMD hasn't been much better. Recent cards like the RX 9070 XT and 9060 XT do bump up to 16GB, but the 8GB RX 9060 XT and even older options like the RX 7600, 6600 XT, and 6650 XT still populate store shelves with 8GB VRAM and tempting discounts. False economy in 2025 It's easy to think you're saving money by buying an 8GB card, but that short-term gain quickly erodes. As games become more demanding and AI workloads become more memory-intensive, you'll end up running into performance walls sooner than you'd expect. That leads to either compromises like dropping settings or resolution, or spending more to replace your GPU sooner than planned. If you're buying a card in 2025, aim for at least 12GB of VRAM, preferably 16GB if you want the system to stay relevant for 3–4 years. Cards like the RTX 5060 Ti (16GB), RTX 5070 (12GB), RX 9060 XT (16GB), and even some budget 7700 XT (12GB) models offer a far better long-term experience, even if they're a bit more expensive. Exceptions to the rule? Despite the growing irrelevance of 8GB GPUs in modern AAA gaming, there are still select scenarios where these cards make sense, provided the buyer understands their limitations. Esports titles like Valorant, League of Legends, and Counter-Strike 2 remain light on VRAM requirements and are designed to run at high frame rates even on modest hardware. For gamers who stick to 1080p resolution and play older or well-optimized games, 8GB cards can still deliver decent results. Similarly, budget-constrained builders, those with less than $300 to spend on a GPU, may find that 8GB cards are their only option unless they opt for second-hand GPUs with higher VRAM but lower efficiency and weaker features. There are also workloads where VRAM isn't the primary bottleneck. Media-centric systems, such as HTPCs or dedicated streaming rigs, can benefit from the video encode/decode capabilities of modern 8GB cards, especially if AV1 support or low power draw is a priority. Small form factor (SFF) builds or compact office PCs often can't accommodate large or high-wattage GPUs, and in such contexts, a compact 8GB card may be the most practical choice. Some users also rely on cloud services like GeForce Now or Adobe's AI-based rendering tools, where the heavy lifting is offloaded to remote servers. In these hybrid workflows, the local GPU serves more as a bridge than a workhorse, making an 8GB card a tolerable, if not an ideal solution. These use cases won't apply to everyone, but they do highlight that there's still a narrow but valid market where 8GB GPUs haven't been completely left behind. Final thoughts The writing is on the wall: 8GB GPUs are no longer a smart buy in 2025. Between games that already exceed that memory envelope and hardware cycles moving quickly toward more demanding workloads, buying an 8GB card today is like buying a smartphone with 64GB of storage. It technically works, but you'll regret it the moment you use it in the real world. If you want to build a PC that lasts, gives consistent performance, and doesn't force you to dial back settings in every new game, skip the 8GB options. It's no longer enough.
Yahoo
26-06-2025
- Yahoo
Leaker believes 'there's a surprise' with rumored new take on Nvidia's RTX 5090 GPU – but if that's a cheaper price, there'll be a sting in the tail
When you buy through links on our articles, Future and its syndication partners may earn a commission. Nvidia could have a new spin on its RTX 5090 graphics card This is rumored to be a new version of the RTX 5090 D for China A leaker says it'll pack a 'surprise', but the truth could be more mundane Nvidia is rumored to be planning a new variant on its Blackwell flagship graphics card, potentially an RTX 5090 DD model – and it could come with a 'surprise' according to one leaker. Tom's Hardware flagged up two separate mentions of this potential GPU (add seasoning with all this) on X, the first of which came from MEGAsizeGPU, who mentioned the RTX 5090 DD and how it will be for China only. So, the theory is that this is a follow-up to the RTX 5090 D (single D, not double D) for China, a different spin on the standard Blackwell flagship which was produced for that country to get around US export regulations. Except the goalposts on those restrictions have been moved, and now rule out the RTX 5090 D itself as too powerful (for AI usage) to be shipped to China. It looks like the idea now is to further cut down the graphics card to again make it feasible to supply to Asia, and call it the RTX 5090 DD. The purported cut-down specs were supplied by another regular on the GPU leaking scene on X, Kopite7kimi, as you can see below. The key changes are a slightly lower-tier GPU chip (GB202-240, rather than GB202-250 as seen in the original 5090 D), and the video RAM has been dropped to 24GB with a narrower 384-bit memory bus. That means a 25% hit on the memory bandwidth compared to the RTX 5090 D, but the core count and TDP remain the same (the latter is interesting, and I'll return to why shortly). Tom's Hardware takes it as read that these two rumors are fully aligned – and they are from what we can see spec-wise – but note that Kopite7kimi has a question mark heading up their post on X, which begins: 'RTX 5090 DD?' However, this is likely referring to questioning the name, and whether this will really be called the RTX 5090 DD, rather than doubting whether this is actually a new spin on a China-only graphics card. Although we should never take anything for granted from the rumor mill, and maybe the latter could be the case. At any rate, the most intriguing bit here is Kopite7kimi's mention of a "surprise" from Nvidia with this GPU. When it comes to the name, yes, making it 'DD' sounds rather implausible, but remember – this is the company that brought you the RTX 4070 Ti Super graphics cards (and similarly clunky bouts of unimaginativeness in the past besides). As mentioned, the surprise is what we really want to know about, and Kopite7kimi doesn't elaborate on what that might be with any kind of a vague hint, even. So, we're left to guess on that score. Tom's Hardware theorizes that the surprise could be further cut-backs to the AI powers of the graphics card delivered via the firmware, and that sounds entirely possible. But equally it sounds entirely dull, and the way the leaker teases that "there's a surprise" feels like it's phrased to stoke a bit of excitement here. A minor chop to AI capabilities hardly qualifies in that respect, but what else could Nvidia possibly have up its sleeve here? For me, this hints that Nvidia is planning a move like cutting the price of the RTX 5090 DD. Or perhaps, most excitingly, offering it outside of China, too? Yeah, okay, that seems unlikely (although as noted, with that question mark, it could be that Kopite7kimi isn't fully sure this is the mentioned Chinese variant). Perhaps more realistically, we could be looking at a fair old bump in clock speeds (hence the TDP staying the same, with the spec being cut down in some key areas). Hopefully, we'll find out soon enough what Nvidia might be up to here, if anything, as this could all turn out to be so much smoke from the rumor mill (or plans from Team Green that never come to fruition). AMD looks like it's losing the GPU war based on new Steam survey, with Nvidia's RTX 5060 Ti proving itself to be popular already Rumored Nvidia RTX 5080 Super specs disappoint some gamers, but I don't think there's anything to worry about with this GPU AMD could be working on a faster RX 9070 Extreme graphics card
GSM Arena
25-06-2025
- Business
- GSM Arena
Nvidia announces GeForce RTX 5050 for $249
Nvidia today announced the smallest member of the RTX 50-series, the GeForce RTX 5050. It will start at $249 and will be available in the second half of July. The RTX 5050 features a new GB207 die based on the Blackwell architecture with 2560 CUDA cores, two thirds of the RTX 5060 that launched in April. It also features 5th generation Tensor cores and 4th generation RT cores like other 50-series cards. This enables it to support features like DLSS4 with multi-frame generation and the upcoming Reflex 2. The RTX 5050 GPU is clocked at 2.31GHz base and boosts up to 2.57GHz. You get 128-bit 8GB GDDR6 memory, making it the only 50-series card so far without GDDR7. Other features include 3x DisplayPort 2.1b and 1x HDMI 2.1b connectivity, 1x 9th generation NVENC encoder and 6th generation NVDEC decoder. The card runs on 130W of power and can take either a single 8-pin connector or 300W or greater PCIe Gen 5 cable. Nvidia has a couple of mostly pointless graphs showing the relative performance of the RTX 5050 compared to older Nvidia cards that make it hard to make anything out but the company claims the new card is 60% faster on average in raster than the RTX 3050 that came out over three years ago. The card also seems on average in the ballpark of RTX 4060 performance from two years ago. The RTX 5050 should be on shelves this time next month. There is no Founders Edition model from Nvidia but the card will be available from all major board partners, including ASUS, Colorful, Gainward, Galaxy, GIGABYTE, INNO3D, MSI, Palit, PNY, and ZOTAC. There is also going to be a mobile version of the RTX 5050, which is shipping today in laptops starting at $999. Source
Yahoo
16-06-2025
- Yahoo
Unreleased Nvidia RTX 3080 Ti 20GB Founders Edition engineering sample sells for $1,999 on eBay
When you buy through links on our articles, Future and its syndication partners may earn a commission. A rare engineering sample of the Nvidia RTX 3080 Ti Founders Edition with 20GB of VRAM was recently sold in an auction on eBay. This specific variant never managed to reach retail, thus making it one of the most unusual Ampere-era GPUs to surface online. According to the listing, the auction concluded with the buyer offering $1,999, which is pretty wild for a high-end graphics card that is two-generations old. The seller included photos of the GPU, showing the original packaging box and a bright green sticker on the GPU that reads, 'Not for sale, for development only.' The seller also confirmed that the unit is in excellent condition and fully functional. However, since it is not a retail product, running games or any 3D workloads would require the use of a third-party driver such as Nvidia-patcher, as there are no official drivers available for this card. This could potentially translate to poor or unstable performance, although that remains unconfirmed. Image 1 of 3 Image 2 of 3 Image 3 of 3 The retail version of the RTX 3080 Ti made its official debut in June 2021 with a starting price of $1,199. It featured 12GB of GDDR6X memory and was essentially an RTX 3090 with half the VRAM and a smaller cooling solution. Introduced midway through Nvidia's regular product cycle, it offered higher specifications and performance compared to the RTX 3080. The first rumors of a 20GB RTX 3080 Ti sparked back in December 2020 when a Gigabyte EEC listing revealed several RTX 3080 Ti 20GB models. This was followed by a firmware anonymously uploaded to TechPowerUp, meant specifically for Gigabyte's range of RTX 3080 Ti 20GB GPUs. Eventually, in September 2021, a few months after the official launch of the RTX 3080 Ti, a cryptocurrency miner revealed in a YouTube video that retailers in Russia were selling Gigabyte branded RTX 3080 Ti 20GB GPUs ranging from $2,733 to $2,837. The individual managed to get hold of a rare Gigabyte Aorus GeForce RTX 3080 Ti Xtreme (GV-N308TAORUS X-20GD), which was purchased from HARDVAR, a retail store in Saint Petersburg. At the time of posting the video, the GPU was priced at $3,067. The video also revealed that the only difference between the two GeForce RTX 3080 Ti models was the memory capacity. The 20GB variant offered a 67% increase in GDDR6X memory compared to the standard 12GB version. However, the higher memory model came with a narrower 320-bit memory bus instead of 384-bit. As a result, the 20GB model had lower memory bandwidth of 760 GBps compared to 912 GBps on the 12GB version. It is possible that the RTX 3080 Ti 20GB was planned as a counter to AMD's high-VRAM Radeon offerings at the time. Perhaps it was intended to target users who prioritized memory capacity for gaming, crypto, or creative workloads. Whether Nvidia changed its mind due to production constraints, shifting strategy, or overlapping performance tiers, this unreleased GPU will likely go on to become a "what-if" in Nvidia's Ampere lineup. Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.
Digital Trends
05-06-2025
- Business
- Digital Trends
3 GPUs you should avoid buying right now
Shopping for one of the best graphics cards is a minefield right now. It's not that the latest generation is bad (that's a different subject entirely), it's that the prices still haven't gone back to normal, despite it being a few months since these GPUs first surfaced. If you have an unlimited budget, then of course, you can buy any GPU — even the mighty powerful RTX 5090. But if you're trying to squeeze out the most performance-per-dollar, there are some GPUs I'd recommend avoiding right now. Recommended Videos AMD RX 9070 XT Adding the excellent RX 9070 XT to this list is something I'm doing with a heavy heart. The GPU received stellar reviews from just about every publication that got to try it, and those who ended up buying it largely seem happy with it, too. I have no complaints about the performance of this graphics card; if anything, it surpassed the expectations that I had for it going into the launch of RDNA 4. Capable of rivaling the RTX 5070 Ti, the RX 9070 XT arrived with a $599 price tag, and at first, it was glorious. Gamers flocked to various retailers, with AMD's Frank Azor sharing a photo of a long line of shoppers waiting outside a Microcenter. The consumer market welcomed the RX 9070 XT with open arms. Perhaps the welcome was overly enthusiastic, though. Demand today for our new @amdradeon cards has been phenomenal. We are working with our AIBs to replenish stock at our partners ASAP in the coming days and weeks. MSRP pricing (excluding region specific tariffs and/or taxes) will continue to be encouraged beyond today so don't… — Frank Azor (@AzorFrank) March 6, 2025 Let me reiterate: The problem with the RX 9070 XT is not its performance. I don't think anyone can find any major fault in it … at its intended $600 price point. But unfortunately, the recommended list price (MSRP) didn't last all too long. Once the initial batch of RX 9070 XTs (and non-XTs) sold out, which didn't take long, the prices started rising. Upon launch, there were many models above MSRP, but not wildly so — and some cards were indeed sold for $599. These days, outside of rare stock drops, good luck finding an RX 9070 XT at MSRP. The cheapest option I was able to find on Amazon costs $848, meaning nearly $250 above the MSRP. At that price point, the RX 9070 XT loses some of its charm; the RTX 5070 Ti costs $900, and with comparable performance but better ray tracing and DLSS 4, many GPU shoppers will choose it over the AMD card. Nvidia GeForce RTX 5060 Ti 8GB This was always bound to happen. Nvidia's RTX 5060 Ti launched in two flavors: One with 8GB VRAM, and one with 16GB. The 8GB model starts at $379, followed by $429 for the 16GB version. At such a small price gap, it was almost a given that most gamers would favor the version with more VRAM, even though they both share the same 128-bit interface (which affects their bandwidth, and that also means their overall performance). The frame rate gap between the two versions of the RTX 5060 Ti might not be as large as it seems, but still, the 8GB version is a GPU you should avoid right now. Unsurprisingly, the RTX 5060 Ti with just 8GB of VRAM sells at MSRP on Amazon. There are models that sell for more, but in general, you can grab it at $379. But, while it can match the RTX 5060 Ti with 16 gigs of memory at 1080p, it's always better to have more VRAM if possible, and the 5060 Ti 16GB wins most benchmarks at 1440p. If you're set on buying the RTX 5060 Ti, you might as well spend $100 more and get the (currently overpriced) 16GB version. But if you want my advice, I'd say wait it out and hunt for a 16GB model at MSRP — they do pop up occasionally. Intel Arc B580 This is another GPU that I am sad to have to mention on this list. The Arc B580 surprised everyone when it launched. Our reviewer called it the '$249 Nvidia killer,' and who would've expected this kind of results from an Intel Arc GPU? Knowing the rocky start that the Arc Alchemist generation had to go through, many were wary of Battlemage, but Intel surprised us in the best way possible. The GPU offered fantastic performance for the inexpensive $250 price point — something which is rare right now, at a time when most GPUs cost well above $350. Unfortunately (for the buyers, at least), the Arc B580 wasn't able to stay at MSRP for too long. The GPU now sells for $359 to $409 on Amazon, and at that price, you're better off getting the new RTX 5060. If there's one thing these GPUs have in common, it's that they're all selling far above the price point they were made for. Solid in their own right, they fail to impress when they're suddenly placed one or two pricing brackets above the one they belong in. But if sold at MSRP? All three are worthwhile picks. If you're currently buying a GPU, my advice is to keep an eye out for models selling at MSRP, and then be quick, because they continue to sell out fast.
