GPU boost 29.487.melt probably trying to override whatever you're doing. Watercooling would inherently crop a smidge off the total power, but maybe 5W at most. You'd probably have to amend the boost table in the bios to get it to stop playing the fool, or just use the voltage slider and cull voltage across the board.
RTX 5090 prices stabilising at around 2500 for the base models (e.g. Palit Gamerock) in stock at OCUK and on the bay. Still above RRP (those base models should be less than 2200 IIRC) but slowly returning.
I realise I’m going to sound like an LTT Stan here, but I’m really not. I’m just tired of seeing the same old attempts at stirring up drama over and over again. Especially when the video that attempts to stir up drama is demonstrably and factually wrong. So, now I have had time to watch h the full video, and it’s a storm in a teacup, just like I thought it was: It basically boils down to: why ask a company rep when you have a team of engineers who can rip these things apart and give you a definitive answer, and you have more than enough money to pay horrific amounts of money to shitty scumbag scalpers for GPUs. You didn’t do any of that, you trusted the word of a PR rep, so now you’re incompetent and spreading misinformation. So… ok… Here’s the contentious claim from MSI: Read that wording carefully: “only the FE cards have the design to run all the pins into a single pad on the PCB”. It’s subtle if you’re not familiar with electronics design, so let me add some emphasis: “only the FE cards have the design to run all the pins into a single pad on the PCB”. That statement isn’t wrong. It’s not a lie, whether by intention, omission, or incompetence. Because all the examples shown in the “Actually Hardcore Overclocking” video - besides the FE - have each pin on the connector going to separate pads on the PCB. That doesn’t violate the PCIe SIG specifications either. The spec says that all pins should be connected to the same polygon. I appreciate that this doesn’t mean much to anyone who hasn’t used PCB CAD software to design their own boards, but what it means is that all pins have to be connected to the same piece of copper on the board. Nvidia have all 12 pins going to 2 pads, both connected to the same power & ground planes; other manufacturers have all the pads going to separate pads on the PCB, and those pads are all wired to the same pair of power & ground planes. So there’s definitely nuance there, and that nuance wasn’t explained in the LTT video, so that specific point is misleading to a lay audience not familiar with electronics engineering. But there was a correction issued, which apparently involved an updated and re-edited version of the video in addition to the pinned comment. The PCIe SIG specs document, by the way, costs $4500 to buy (far as I can tell) if you’re not a PCIe SIG member, so I can’t verify the wording myself. I’m trusting that “Actually Hardcore Overclocking” is correct in the wording they use, despite being wrong about MSI’s statement being a lie. Meanwhile: 12V-HPWR is still a stupid design, GPUs still cost a fortune, no one can even buy one unless they’re prepared to pay god knows how much over the odds to immoral and unethical profiteering arseholes, and people never tire of making content that dunks on LTT’s every misstep, even when those missteps might be imagined. So… In summary, a claim from a PR rep was misleading, it was explained in a way that could be misleading, and when this was pointed out a correction was issued in the form of an updated video. It’s yootoob drama… again… that’s pretty much all it boils down to.
D'you know which specs doc it is? 'cos I've got... PCIe Base Specification in various revisions, PCIe Card Electromechanical Specification in various revisions, PCIe External Cabling Specification, PCIe 225W/300W High Power Card (bless!) Electromechanical Specification, the PCIe Connector High Speed Electrical Test Procedure, the PCIe to PCI/PCI-X Bridge Specification, the M.2 specifications...
PCI-E SIG you say? A friend I was talking to recently was saying that his company is a a member of said organisation.
It was already well established long before either video released that only the Asus Astral has any monitoring and no card has any load balancing or safety shut off. So yes, it is indeed just dumb youtube drama over a statement from a pr rep.
MSI Ventus 3X RTX 5080 in stock at Scan for £1040. Looks like we're intermittently but mostly back to RRP.
That’s more of a question of the physical dimensions than the GPU model/generation. It can be worth having a support even for a 2-slot card if it’s long enough. My 4070S, for example, is small enough to not benefit from one, but my previous 3070Ti would sag alarmingly without a support.
Sapphire 9070 Pulse, it's only a two fan. Not sure why the non-XTs are even three fans, this thing is blummin cool as it is
Been looking of the 3 fans ones as my case is pretty big. I have a 2 fan msi 6750xt and it just looks silly in the case
The Pulse is 280mm long, 2.5 slot. Not sure about the others. Yeah I can understand, my case is a matx one so it looks fine even though it is big enough to fit an XT
You don't have to, a lot of the cards have some sort of bracing that ties back to the prear pcie cover, but if a slightly wonky GPU upsets your eyes, they're a quick, easy and cheap fix. There's plenty of adjustable screw type ones on Amazon for a few quid.
