Search results

I don’t know why but feel the need to point out that this card is literally just a 6650xt made on 6nm instead of 7nm.

Well I’ll be damned. They cut the shader count down per EU and are trying to make it back up through dual issue? They should ask AMD how that works out without having an absolutely brilliant compiler to catch all the dual issue opportunities.

I’ll happily bet a hundred dollars that it’s literally just the recompilers made to run CUDA code on AMD or Intel just combined into one package.

I’ve not seen any evidence that Battlemage will only have half the shaders per Xe core. You’re the first person I’ve seen making that claim.

You clearly don’t get what I’m saying but that’s cool. Qualcomm is going to put their Oryon cores into server chip or possibly even a CPU/GPU combo chip like everyone is coming out with right now. Their Adreno GPU architecture is basically based on GCN with optimizations towards running on low...

Isn’t AmpereOne just a neoverse core like Ampere Altra was, but just a newer version?

He’s obviously talking about the Linux community when he says “cult”, not a hardware vendor. Also the Grace CPU uses off the shelf ARM cores that are much less potent than Apple core architectures or the new Qualcomm core architecture. Grace is literally just a slightly newer Neoverse core than...

I can’t understand why people can’t see that the Zuckerbergs and Musks, and Bezos(s?) and Altmans (maybe even the Huangs and Sus) of the world are absolutely every bit as dangerous and corrupt as the Rockefellers or Carnegies were.

What? The AI/overall compute performance was the only thing Alchemist was already good at.

Isn’t it also AMD?

Intel releases this every month. Lots of companies do a rollout on patch Tuesday.

Definitely. It’s always hard to get huge amounts of hardware to scale properly over a large area. Hell Cray doesn’t even make its own products anymore but it’s still a viable business just from having experience getting these things operable.

Honestly if they can hit 4070 performance at like $300, I’ll buy them for my friends that don’t have a GPU. The a750 might’ve sucked overall but not for $180 it didn’t.

Can’t we just limit business energy consumption instead of building tons of new power plants? In the long term, AI is just going to make jobs disappear anyways.

Laptops generally have hard limit on RAM set in the bios just like motherboards have a hard limit of what is considered “supported”.

Well first of all, nobody could prevent Valve from setting up a cloud gaming service, but it’s not likely as it would present a complete shift in business model. As far as Strix Halo, I think you’re overblowing the cost. 40 CU and 16 cores should fit into 400mm^2 fairly easily on TSMC 4nm. 8...

Good ol Ponte Vecchio

Uhmm…anybody got some bleach I can pour in my eyes?

Everything I’ve seen says it’s l5x-8000, rather than 8533

You’re right. It’s 100% the PCIe controllers that heat up rather than the traces themselves. Also a lot of the heat is coming from device controllers running at higher speeds to match the PCIe controllers. If one actually reads the article it’s pretty clear that’s the sort of thing they’re...

In some use cases PAM4 encoding allows reduced clock speeds because it transmits more data per clock cycle. Like I know in Nvidia Ampere GPUs that used gddr6x memory, they actually ran at lower clock speeds but used PAM4 encoding to increase data transmitted per cycle by 2x to end up at higher...

I’m quite sure the MI300 is going mostly to private businesses that don’t want to shell out Nvidia prices. Why would a cloud provider dealing with LLM training want to sell anything other than Nvidia currently? As someone who’s played around with ML a fair amount, Nvidia has such a performance...

I also work for a Fortune 500. I have zero ideas what our numbers are. I’m quite sure I’m counted as a contractor though.

Then what’s even the point if there’s no OS? You can use microcontrollers as a cpu though. Look at the Raspberry Pi Zero. It’s sold as a microcontroller. If this thing literally only blinks LEDs then I’ve never seen something this stupid.

I guess it should at least run Linux since it’s risc v

If you think the average consumer is ready for multiple CPUs, I get the impression you’ve never used a multiple CPU server. It would make much more sense to just give consumers access to more cores on consumer CPUs. If Intel or AMD would sell a consumer CPU with like 32-48 e cores and 64 pcie...

Just for the record, Nvidia has 30,000 employees. Not 26,000. The AMD number is right. Considering Nvidia concentrates EVERYTHING on GPGPU computing and Radeon is a tiny shithole in the closet of the CPU department at AMD, it’s kind crazy AMD gets as close as they do in overall GPU performance...

Isn’t getting 10% of the sales Nvidia is getting basically a straight up 9% market share gain from zero? That’s actually encouraging considering Nvidia was the only show in town.

You keep claiming 100% of cases were shown to be user error but that was only ever written by some online magazine. We don’t know if that’s really true. In fact, I highly doubt it. No product in the world has ever had 100% of failures caused by user error..

Dude the connector is a poor design. I’ve got a 4080 and it absolutely doesn’t give a good firm click like the old 8 pins.

No dude. Don’t pretend CPUs were exploding and don’t pretend it was anything like as widespread as this. I’m currently running two Raptor Lake desktops and a Zen4 laptop and this is a much bigger problem as far as I’m concerned.

I’m fully aware of how chips work. Also, with proper cooling the 13900k is easily capable of 5.8 all core. Now sure if you’re running Linpack or p95 that will be nearly impossible to keep cool but it’s quite easy to get through Cinebench runs that way. But yeah it’s clear Intel needs to bin them...

This many people don’t just “fail to plug it in properly” for no reason. Even though I haven’t had the issue since I only have a 4080, I fully recognize the plug is NOT a good design with a nice firm click like it should be.

As far as I’m concerned SSDs have gotten cheap enough to just store everything on NAND instead of HDDs with sort of NAND based caching like you’re implying.

Yeah any chip with 40% of its SMs disabled was indeed likely to be in pretty bad shape to start with.

Probably because that memory controller is in fused off area of the chip

As someone who’s had a 13700k and 13900ks. It’s pretty clear to me, that the 5.8-6.0 frequency range just requires tighter binning than they’re applying if they want good stability. Any 13th gen and up k chip is good at 5.5-5.6, but for the chips that boost higher than that, they need to apply...

I’ve been trying for years to figure out what was left activated in the 2060 ko that raised it performance above the 2070 in some productivity apps. It wouldn’t have been ROPs, as they’re number is fixed per SM active.

They only risk Huawei sanctions if they openly work with the CCP in a militarized fashion.

Fair enough. From what I’m seeing, they should be fairly close to n100 level now if they’re floating between Skylake and Kaby Lake performances. The last n100 scores I’ve seen were pretty close to 6500/7500 non-k performance.