Arc B580 power consumption: Battlemage GPU seemingly improves upon Arc A580 but RTX 4060 still reigns supreme

[Redaktion]

The Intel Arc B580 Battlemage GPU has debuted to glowing reviews with the reviewers praising the card's price/perf. However, despite the Arc B580's raw performance, the GPU lags behind the RTX 4060 in power consumption.

https://www.notebookcheck.net/Arc-B580-power-consumption-Battlemage-GPU-seemingly-improves-upon-Arc-A580-but-RTX-4060-still-reigns-supreme.932120.0.html

[Just Checking]

Just wondering, does anyone care about power consumption apart from those who already have a pc with low power supply than 4070 need. Also, many leave a bit of headroom for future upgrades.

[Check no further]

Doesn't seem like it does it?

Honestly, I think people do care especially in the EU where power is more expensive and a significant cost.

The only reasoning for why I can see it's selling as well as it is, is mostly due to combination of not much initial stock at launch and basically enthusiasts & reviewers getting this card.

Think the vast majority are waiting for navi 44 and RTX 5060.

Hoping Xe3 is more competent in performance per watt area because it seems in nearly every other area intel has caught up, besides maybe dx9 performance.

[GeorgeS]

As others have mentioned...

For many years (even DECADES) GPU cards have required 100-300W (or more) of power.

Only the 'lowest end' of the low end PC's may not have 300-400W of power supply 'budget' available to power a single GPU card.

So given the now possible THREE choices of ether a: AMD, Nvidia or Intel GPU card a perspective buyer might look at cost, power budget AND driver support before deciding on which companies product to buy into.

Personally I'd pick an AMD or Nvidia product over Intel ANY day of the week. :)

[A]

Just wondering, does anyone care about power consumption apart from those who already have a pc with low power supply than 4070 need. Also, many leave a bit of headroom for future upgrades.

It can vary based on electricity prices, though most people don't really pay much attention as people generally look at upfront cost than cost over time. Most people in general see higher energy consumption = more heat if anything

That said, with trend of crypto and AI, more focus has been put on power consumption because before you only used power when gaming. But these workloads many run even when someone isn't there on the computer and unlike gaming which has varying consumption based on what you are running, Crypto and AI easily push your card to the max for speed. At that point the bills become more noticeable.

[@man_daddio]

I think the only thing that matters here is price.
If you're getting 20% more performance in some games at a lower price than the competition then I think that that would be more valuable than worrying about 40 more watts which isn't going to matter unless you're gaming 12+ hours a day.

[RobertJasiek]

I think the only thing that matters here is price.

Uh, no. Compatibility to used software is essential.

[Worgarthe]

I think the only thing that matters here is price.

Uh, no. Compatibility to used software is essential.

Why would there be no compatibility? It's not an issue. Video by Level1Techs: The Intel Arc B580: Midrange GPU Magic!

[RobertJasiek]

Nvidia CUDA, CuDNN and TensorRT libraries and softwares using them do not run on the cores of Intel or AMD GPUs.

[Worgarthe]

Nvidia CUDA, CuDNN and TensorRT libraries and softwares using them do not run on the cores of Intel or AMD GPUs.

This is specifically a 1440p gaming GPU in a sub-US$300 market, it was announced like that, it's advertised like that. You currently can't buy anything on the market that comes anywhere near that level of gaming performance for so little money. RTX 3060, RTX 4060, RX 6600, RX 7600, RX 7600XT, all of that is slower and more expensive (well not the 6600, that one is a bit less expensive but still 40% slower).

You can buy a gamepad controller and use it to move your mouse pointer around, it will work but it's not made for that type of usage. Same with this GPU.

[A]

Nvidia CUDA, CuDNN and TensorRT libraries and softwares using them do not run on the cores of Intel or AMD GPUs.

This is specifically a 1440p gaming GPU in a sub-US$300 market, it was announced like that, it's advertised like that. You currently can't buy anything on the market that comes anywhere near that level of gaming performance for so little money. RTX 3060, RTX 4060, RX 6600, RX 7600, RX 7600XT, all of that is slower and more expensive (well not the 6600, that one is a bit less expensive but still 40% slower).

You can buy a gamepad controller and use it to move your mouse pointer around, it will work but it's not made for that type of usage. Same with this GPU.

Things are a bit more complex than that, some games have features that require certain gpus like for example particle physics. On top of that, game developers tend to do most testing and optimization on most popular gpus so you are likely to have less glitches and better performance.

There are also things outside of just gaming, like productivity that use these features. Then there is of course AI

Not saying they won't get there with time, but it is what people consider

[Worgarthe]

Things are a bit more complex than that, some games have features that require certain gpus like for example particle physics. On top of that, game developers tend to do most testing and optimization on most popular gpus so you are likely to have less glitches and better performance.

That's true, but the fact is that in all tested games (sample of 200+), by all trustworthy and relevant channels such as Gamers Nexus, Level1Techs, Hardware Unboxed etc., in all but one game the B580 was faster than all other previously mentioned GPUs. So performance is there, drivers and optimisations from Intel are there, and the product is a budget monster for 1080p and 1440p gaming.

There are also things outside of just gaming, like productivity that use these features. Then there is of course AI

Of course, that's true, but they (Intel) have never said a single thing about AI/ML and this card. I mean if one wants to do such things they probably won't gamble with this card at all, they will most likely opt to go with the safe green-coloured route. Then again, plenty of gamers want to just game and they don't care about anything else, especially not if they have more than capable other machine for productivity (say M3 or M4 based MBP). With this card you can build a gaming beast for less than $700 total.

[4n0n]

What about idle power draw? The Intel Arc series seems to have pretty high power consumption on idle iirc.

[SBP]

Nvidia CUDA, CuDNN and TensorRT libraries and softwares using them do not run on the cores of Intel or AMD GPUs.

scale and zluda?