
Gaming Hardware News, July 23 2026: RTX 50 Super On Hold, Zen 6 Arrives
This week the number that explains everything since June finally surfaced: a 3GB GDDR7 chip costs $60-70 against roughly $20 for a 2GB part — nearly triple — and delivers only 50% more capacity in return. Run that math and the RTX 50 Super cards aren’t shipping because they can’t ship at a price that makes sense. That, plus the uncertain future of Intel’s Arc line, says a lot about where your money should go right now.
Nvidia puts the RTX 50 Super launch on hold over GDDR7 memory pricing
The RTX 50 Super cards are reportedly built but held back over memory costs: a 3GB GDDR7 chip runs $60-70 against ~$20 for a 2GB part, for only 50% more capacity. It affects the RTX 5070 Super (18GB), 5070 Ti Super (24GB) and 5080 Super (24GB), with no confirmed date (Tom’s Hardware, VideoCardz).
What it means for your next purchase: let’s do the arithmetic, because that’s where the value is. An 18GB RTX 5070 Super needs six 3GB chips. At $60-70 each, memory alone costs $360-420. The standard 5070, with 12GB across 2GB chips, uses six at ~$20: $120 of memory. That’s $240-300 in additional component cost, before manufacturing, board partner margin, distribution and retail markup. By the time it reaches a shelf, you’re looking at $300-400 on the final price.
An RTX 5070 Super would cost roughly what an RTX 5080 costs this week (~$1,249). That’s the point at which the product stops existing commercially — nobody buys a mid-range Super at last year’s high-end price. It isn’t a scheduling decision or a marketing play; the product simply doesn’t add up.
Three practical consequences:
- Stop waiting for the Supers, permanently. This isn’t a delay measured in months. It’s a product whose viability depends on GDDR7 getting cheaper, and nothing suggests that happens in 2026.
- High VRAM is a luxury for a while. This cuts against what all of us have been saying for years (“buy extra VRAM to future-proof”). At these costs, paying for VRAM you won’t use today is throwing money away. I’ve revised my thinking in how much VRAM you need in 2026: buy for the resolution you actually play at.
- Current cards just became more valuable, not less. With no refresh coming, today’s 5070 and 5070 Ti hold their position far longer than usual. Buying now doesn’t leave you with “the old model” in six months. See best value graphics cards.
AMD officially debuts Zen 6 with the 256-core EPYC “Venice”
AMD unveiled Zen 6 on July 22-23: EPYC “Venice,” the first chip on TSMC’s 2nm node, with up to 256 cores (+33% over “Turin”) and a claimed 70%-plus improvement over Zen 5. It’s a server part; desktop Ryzen is expected around CES 2027 (Tom’s Hardware, TechPowerUp).
What it means for your next purchase: I covered this last week and the conclusion hasn’t moved — server chip, eighteen months to desktop, and the 70% doesn’t translate to games.
The one thing I’d add is about the 2nm node, because it does reach you: new nodes usually mean constrained production early on. If desktop Zen 6 inherits that, the first Ryzen parts of 2027 arrive with thin stock and high prices, exactly what’s happening to GPUs now. The good buy won’t be CES 2027; it’ll be well after. Combined with AM5 support through 2029, the practical conclusion stands: buy AM5 now and upgrade within the platform. See best CPU for your budget.
Intel’s gaming Arc GPU future is thrown into question
Intel has reportedly cancelled the dedicated gaming GPU based on Xe3P “Celestial.” The next possibility is Xe4 “Druid” in late 2027, with no guarantee of a gaming variant. Separately, an Arc B770 (Battlemage, 32 Xe2 cores, 16GB+) has leaked (WCCFTech, Club386).
What it means for your next purchase: this is the worst long-term news of the week and deserves an honest reading. Intel was the third competitor applying price pressure at the low end — a 12GB Arc B580 at $249 is precisely the product that forces Nvidia and AMD to respond. If Intel exits dedicated gaming graphics, that pressure disappears, and in an already-strained market that’s bad news for every buyer, whether or not you’d ever own an Arc.
That said, don’t overreact:
- If you already own an Arc Battlemage card: nothing changes today. Intel is still shipping drivers and your card works as it did yesterday. Worry when updates stop, not when a future architecture gets cancelled.
- If you were about to buy a B580: the real risk isn’t current performance, it’s driver support three to four years out. Arc cards depend on driver updates far more than Nvidia or AMD, because their performance in older games has improved specifically through patches. If Intel scales back, that engine stops. I’d only recommend one if the price clearly offsets that risk — and above MSRP, it offsets less every week.
- On the leaked B770: a leak isn’t a launch. Don’t plan around an unannounced product, least of all from a vendor scaling back its commitment.
ASUS ROG launches its Tandem OLED PG27UCWM and PG32UCWM monitors
ASUS announced the ROG PG27UCWM (26.5”) and PG32UCWM (31.5”) on July 15, with Tandem RGB OLED panels and a TrueBlack Glossy finish, running dual mode: 4K at 240Hz or Full HD at 480Hz, with DisplayPort 2.1a UHBR20 and Dolby Vision. No pricing announced (ASUS Pressroom).
What it means for your next purchase: the dual-mode idea is genuinely clever — one panel for games you want to look good and games you want to win. But someone should say the part the press release doesn’t: you need a GPU capable of driving 4K at 240Hz for any of this to matter. At this month’s graphics card prices, that GPU costs more than plenty of complete systems. A monitor like this on a mid-range card is money sitting idle.
Tandem OLED is also new, and these are among the first panels to market. First-generation anything tends to arrive at launch pricing with a firmware revision or two ahead of it.
Update, August 6, 2026. There was no price when we published this. There is now: €1,249.90 for the PG27UCWM and €1,359.90 for the PG32UCWM, initially in Germany, Austria and Switzerland (ASUS Pressroom). That confirms the point above with room to spare: at that price, the monitor only makes sense if you already have — or are about to buy — a GPU that can drive 4K at 240Hz. Add the two together and this costs more than most of the complete builds on our lists.
My recommendation is simple: if you’re building now, the monitor is the last place your money should go this month. A good 1440p panel at 144-180Hz costs a fraction and leaves budget for RAM and a GPU, where the memory crisis is actually squeezing. Options in best gaming monitor and best curved gaming monitors. And if you’re still tempted, wait for pricing — with none announced, there’s no decision to make.
What I’d do this week
| Your situation | What I’d do |
|---|---|
| Waiting for the RTX 50 Super | Cancel it. They won’t ship at a sane price. |
| Need a GPU now | Buy in your current tier. These cards will stay in catalogue. |
| Planned to buy “extra” VRAM | Not anymore. Buy for the resolution you play at. |
| Considering an Arc B580 | Only if price offsets four-year support risk |
| Waiting for Zen 6 | Buy AM5 now. Supported through 2029. |
| Tempted by the ROG Tandem OLED | Wait for pricing. Sort GPU and RAM first. |
Sources
The GDDR7 cost figures, the Xe3P cancellation and the Arc B770 come from leaks and supply-chain analysis, not official statements. The Zen 6 and ASUS announcements are official.
- Tom’s Hardware — 3GB GDDR7 memory cost
- TrendForce — memory pricing in Q3 2026
- VideoCardz — RTX 50 Super on hold
- Tom’s Hardware — AMD Zen 6 and EPYC “Venice”
- TechPowerUp — 6th Gen EPYC “Venice” announcement
- WCCFTech — Xe3P “Celestial” cancellation
- Club386 — Celestial cancelled, Arc’s future in doubt
- ASUS Pressroom — ROG Swift OLED PG27UCWM and PG32UCWM