• expired

Gigabyte AMD Radeon RX 9060 XT 16GB $826.85, Intel Arc B580 12GB $550.85, Nvidia Gforce RTX 5060 Ti 8GB $746.85 & More @ PB Tech

60
Gigabyte AMD Radeon RX 9060 XT 16GB $826.85, Intel Arc B580 12GB $550.85, Nvidia Gforce RTX 5060 Ti 8GB $746.85 & More @ PB Tech
Go to Deal
Huntakillaz on 14/08/2026 - 09:03 pbtech.co.nz (366 clicks)
Last edited 14/08/2026 - 22:46 by 1 other user
PBGAMER26

Some decent deals (in todays market) atm if you need one, be quick with gpu's Nvidia announced price hike and their GPu's went up 10-30% early this week, AMD to follow but seems it hasn't gone up yet for most cards, Intel may follow as well.
Motherboards expected upto 50% price hike too (https://www.techpowerup.com/351339/asus-gigabyte-and-msi-to-…)

Note: Intel GPUs (and certain other PC parts) come with free games: https://www.pbtech.co.nz/promotions/intel-bonus-tomb-raider-…

$826.85 - Gigabyte AMD Radeon RX 9060 XT Gaming 16GB Graphics Card
https://www.pbtech.co.nz/product/VGAGBA190621/Gigabyte-AMD-R…

$550.85 - Intel Arc B580 12GB GDDR6 Graphics Card
https://www.pbtech.co.nz/product/VGAINT1580/Intel-Arc-B580-1…

$527.85 - ASRock Intel Arc B580 Challenger 12GB OC GDDR6 Graphics Card
https://www.pbtech.co.nz/product/VGAASR02580/ASRock-Intel-Ar…

$458.85 - ASRock Intel Arc B570 Challenger 10GB OC GDDR6 Graphics Card
https://www.pbtech.co.nz/product/VGAASR02570/ASRock-Intel-Ar…

$746.85 - NVIDIA GeForce RTX 5060 Ti 8GB Graphics Card
https://www.pbtech.co.nz/category/components/graphics-cards/…

Computing AMD Asrock Computer Component Intel Intel Arc B570 Nvidia Video Card

Related Stores

PB Tech
PB Tech

Comments

Old
Search through all the comments in this post.
  • danvelopment on 14/08/2026 - 09:22

    I don't know about gaming, but Intel ARC GPUs are the ultimate for video (encode/decode/transcode/format support) on a price/performance basis.

    I use an A380 for my JellyFin server.

    That being said, for video I'd buy a last gen Intel GPU as they're better value than current gen.

    • 1
    • Huntakillaz on 14/08/2026 - 09:26

      They're pretty good on gaming for the most part just expect a bit more hiccups on occasion vs AMD/NVidia as their drivers aren't as mature

      Also they run better on AM4/Intel 1700 or newer sockets as they require rezsize bar support to run better
      https://www.reddit.com/r/IntelArc/comments/1hw6iao/what_is_r…

      Gaming Reviews:
      Intel Arc B580 Review, The Best Value GPU! 1080P & 1440p Gaming Benchmarks
      https://www.youtube.com/watch?v=aV_xL88vcAQ

      Huge Arc B580 News! Intel Fixes CPU Overhead Problem
      https://www.youtube.com/watch?v=gfqGqj2bFj8

      Intel Arc GPU Driver Problems Revisited: 2025 Arc Graphics Driver Review
      https://www.youtube.com/watch?v=fhECKZQI_Y8

      • 1
      • Everettpsycho on 14/08/2026 - 09:31

        Some 8/9 gen Intel motherboards have rebar but I heard there was still issues with overhead on the Intel drivers.

        I'm always very tempted to pick up a B580 but still havent got around to doing my dell workstation to new motherboard transplant to check if rebar works on it.

        Interesting time in the gpu market at the moment though, the Linux push is helping AMD along, Nvidia still has the performance and cuda and Intel is priced aggressively. I hope they don't drop out the market like it's rumoured as 3 players will hopefully spice things up when components finally come back to remotely sensible pricing.

        • 0
        • Huntakillaz on 14/08/2026 - 09:36

          Yup Intel hopefully stays in the game, they've got a good foothold atm and just need to iterate more now, especially if the can continue to offer it for lower prices (lower profits) than AMD/NVIDIA equivalent
          We're also seeing a Chinese GPU emerge, though it's more closer to RTX 3060 performance in 5yrs the landscape will be quite different

          Also for anyone wanting to see performance
          Intel ARC B580 – Worth Buying in 2026? (30 Games Tested)
          https://www.youtube.com/watch?v=i3dgASI-7EI

          IMO Bang for buck (issues & slight future proof included) atm is 9060XT 16GB, followed by Intel B580 12GB

          • 0
      • AleBeerenhoff on 14/08/2026 - 09:34

        I switched from a RTX3060 to a B580 and have been very happy with the price/performance.

        • 1
    • stardust on 14/08/2026 - 14:29

      i am running jellyfin on my 1st generation mobile igpu iris xe, 11th gen intel indgrated graphics , its flawless it picks up intel quick sync on debian / mx linux docker can easily push 4 + 1080 stream easily the bottle neck is my wireless bridge lan .

      • 0
    • felixfurtak on 17/08/2026 - 10:42

      A380 for a jellyfin server. That's crazy over specified. I'm running an N150 and it runs multiple 2160p HDR streams flawlessly even transcoding. And that's running as a lxc on proxmox, with open media vault, home assistant vms running at the same time.

      • 0
      • danvelopment on 17/08/2026 - 11:01

        I was going to go for an A310 but the A380 was cheaper.

        And it's not a performance thing, the A310/A380 is the lowest end PCI-E GPU money can buy with AV1/h265 support.

        I have a Tesla P4 in the same machine, but it's only good for h264.

        That being said I could have half a dozen streams coming off my server, it's not just local (in fact it's not even local to me, but physically located a few streets away).

        Nvidia Decode Matrix:
        https://developer.nvidia.com/video-encode-decode-support-mat…

        h265 didn't get full support until Turing, and no AV1 until Ampere. You don't get "low end" Ampere and later chips, nor has there been a low end AMD GPU in a long time (and their decoder/encoders are inferior to Intel/Nvidia).

        Also your N150 is processor limited, it would struggle to run a dozen VMs and hundreds of GB of RAM.

        • 0
        • felixfurtak on 17/08/2026 - 11:33

          Fair enough I guess as long as you don't care too much about power consumption.

          But any 11th gen or above intel cpu with Xe igpu will decode av1/h.265 just fine. Even the lowest end ones.

          • 0
          • danvelopment on 17/08/2026 - 12:59

            @felixfurtak: I don't see why power consumption would be a talking point, they don't idle at their full load, and their full load is 75W or less.

            A whole bare metal machine running an 11th gen Intel CPU would consume more power than a VM with an A380 passed through.

            Plus you'd probably need a monitor to manage it, and a keyboard and mouse, and be physically present to manage the BIOS and/or change its power state.

            • 0
            • felixfurtak on 17/08/2026 - 13:10

              @danvelopment: It makes quite a difference it you're running 24/7.

              My entire N150 media server idles at ~4-5W. That's with 4 SSDs, running tailscale, open media vault, jellyfin, home assistant on proxmox.

              Desktop setups are of course much more powerful, but idle a lot higher. You'll be lucky to idle at 100W with your setup which would be 2.4kWh per day or ~900kWh per year

              • -1
              • danvelopment on 17/08/2026 - 13:19

                @felixfurtak: What does that have to do with anything? I said I have an A380 for JellyFin and now you're talking about the whole machine as if you know anything whatsoever.

                I don't have a desktop setup. I run a server, part of which literally means I don't need a desktop setup (my laptop acts as a thin client for an array of desktop VMs, for one thing). I can safely say my power consumption of the server, whilst high on its own, is lower than the total power consumption of the devices I would have if I didn't have the server, plus the loss of many other functions that server is performing.

                Your argument is stupid, I added an A380 to my machine because it was the single most cost effective method to achieve my goal, and it does a great job. In doing so, it uses less power than your proposed alternative.

                You should get in touch with Facebook and tell them it would be way more cost effective to run their servers on N150s. They're just THROWING money away hosting their machines on servers in datacentres. Zuckerberg should build an N150 box and keep it in his lounge.

                • 0
                • felixfurtak on 17/08/2026 - 13:34

                  @danvelopment: Apologies. I thought you were just running a media server. It's over specified for a media server but if you're using it for other stuff fair enough. Just be aware of your power consumption leaving it running 24/7 that was my only point.

                  For 1-4 users a N100/N150 is fine for a general jellyfin media server. Anything else is overkill https://jellywatch.app/blog/best-gpu-for-jellyfin-transcodin…

                  • 0
  • Donnycrane on 14/08/2026 - 10:33

    I recently splashed out on a 5070ti to replace my 3060ti. Justified it for editing but haven't noticed any difference hahaha! Oh well, at least my boys PC got an upgrade with the old card :)

    • 1
    • Huntakillaz on 14/08/2026 - 10:34

      Hope it was before the $200+ price rise this past week.

      • 0
      • Donnycrane on 14/08/2026 - 10:39

        Hah yeah a couple months ago and man that hurt. The old wait for prices to go down strat has not served me well this generation.

        • 0
        • Huntakillaz on 14/08/2026 - 10:43

          Yeah i was looking to get one to replace my 10GB 3080 in my 2nd pc for VR, but now I'll just wait it out and see what the next gen cards price come out as, hopefully there is at least limited stock at the lower msrp prices then it'll be day one buy this time again.
          Snagged a 9070XT on launch so I'm good atm.

          • 1
  • ace310 on 14/08/2026 - 11:59

    Any good gpu to run AI locally? Don't want to go too far or need really advance capability. Frankly, just want to start/learn something. Also, I have
    ASRock Z370 Extreme 4 ATX Motherboard with Intel i3 8100 and 24gb ram. I might have to upgrade power supply (500w currently).

    • 0
    • Huntakillaz on 14/08/2026 - 12:36

      You'd want one with lots of Vram so Nivida 5060TI 16GB > AMD 9060XT 16GB > 5070 12GB > Intel B580 12GB
      Best bet would be to ask on AI reddit threads
      https://www.reddit.com/r/LocalLLaMA/comments/1vj5wel/buildin…

      • 1
      • Everettpsycho on 14/08/2026 - 12:43

        Question will be what's more important, cuda cores or ram? Would you be better with 8gb from Nvidia or 12gb from Intel?

        For the AI I'm dabbling with at work cuda is a must and it will not work on anything else. Annoying as Nvidia know it and price accordingly.

        • 0
        • squareeyes on 16/08/2026 - 13:32

          VRAM. Any modern GPU can do the inference, it's the VRAM you need to store the model - otherwise it spills to system RAM and becomes unusably slow.. And if you want to do anything more than turn based chat, you need 50% again on top of the size of the model itself (e.g. if you have a codebase) or want longer conversations. The context/KV cache gets exponentially bigger the more you feed it, and you want all that in VRAM too.

          • 0
    • tilted on 15/08/2026 - 13:45

      If youre just wanting to run the models, b580 is a good start, 12gb vram will be enough to run a 7b-13b model without the need for quantization. If you're getting into fine tuning and pushing it, Nvidia ecosystem is the easiest. Vram is basically a constraint where if no memory, its a hard stop whereas more cuda cores = more fast if you have the vram to work with.

      • 1
Login or Join to leave a comment
Login Join