Intel Arrow Lake T-Series CPUs show tiny improvement over prior-gen — 65W and 35W Core Ultra 7 265 and Ultra 7 265T CPUs are only 5% faster than Raptor Lake in CrossMark

Core Ultra 200S CPU
(Image credit: Intel)

Intel is readying budget 65W and 35W versions of its Arrow Lake CPUs under the Core Ultra 200 non-K and T lineups, which are expected to arrive by CES 2025. Early benchmarks show that the Core Ultra 265 family delivers up to 5% better performance than its Raptor Lake counterparts in CrossMark per momomo_us on X, which is relatively lower than expected.

The Core Ultra 7 265 is the direct successor to the Core i7-14700. It features 20 cores (eight P + twelve E) and 20 threads and should have a 65W TDP, which can reach 182W (performance preset PL2) on some boards. The Core Ultra 7 265T retains all these specifications but lowers the TDP to 35W, making it an excellent choice for mini-PCs and SFF builds.

BAPCo's CrossMark repository shows that both CPUs have been tested multiple times using different builds - so we'll average the results for our comparison. The Core Ultra 7 265 and 265T amassed 2,120 and 1,844 points, respectively - landing both CPUs 1% and 5% faster than their predecessors. Compared to AMD's equivalent, Ryzen 7 7700, the Core i7-14700 was already 12% faster, and Arrow Lake extended this lead to 13%. Still, this delta is likely the cause of architectural differences between Intel and AMD.

Swipe to scroll horizontally
CPUFamilyAverage Overall Score
Core Ultra 7 265Arrow Lake2,120
Core Ultra 7 265TArrow Lake1,844
i7-14700Raptor Lake Refresh2,099
i7-14700TRaptor Lake Refresh1,752
Ryzen 7 7700Raphael "Zen 4"1,870

While these results aren't fascinating, CrossMark is one of many benchmarking utilities, so this test isn't enough to draw any conclusions. For context, the 65W Core Ultra 9 285 was recently able to keep pace with the i9-14900K and Core Ultra 7 265K unlocked CPUs in a leaked Geekbench test.

Given that Arrow Lake scales particularly well at below 100W, it is bound to give AMD serious competition in the budget segment. However, we should mention that Intel's sub-10W (N-series) CPUs will stick with Alder Lake since Arrow Lake is expensive to manufacture and package.

Nonetheless, the Core Ultra 200U/H/HX/T/non-K series will likely be unveiled at CES alongside other exciting products from Team Red and Team Green. LGA 1851 adoption should increase in the first quarter as cheaper H810 and B860 motherboards are also coming.

Hassam Nasir
Contributing Writer

Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he’s not working, you’ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun.

  • ThisIsMe
    In the end, a win is a win and an improvement is still an improvement. Maybe that comes in the form of better efficiency (which “everyone” has been begging Intel for) or a less than 10% performance gain. It’s still better and in no way should be “cause for concern” unless you’re just being dramatic.

    The desire for the next big product release with the hopes of upgrading before others in order to boast about it is understood and expected of those with gamer-focused mentality. However, what about all the people who just need a new computer? Whether it’s their first PC, they need an additional one, the one they had became unusable, or whatever, I’m sure they’d be happy buying a newer better CPU for about the same cost as the previous models. Doesn’t really matter what the exact improvements are tbh. Why would anyone want an inferior product for the same cost? Besides, new product releases usually cause prices of older products to drop some amount, thus creating the opportunity to not spend as much on something they would have had to buy anyway since that was the best option before the new releases. Then there are those that only need to replace a few parts in their computer because they’re handy enough to know how to do that. Why would they want to be limited only to what has been available simply because the newer stuff is only a little better in some areas?

    Anyway, all that to say these articles are getting extremely dramatic. It’s approaching the level of comment trolls tbh. At worst, the random internet findings presented here are something to consider when comparing the offerings to other available products, if the findings are even found to have merit.
    Reply
  • usertests
    However, we should mention that Intel's sub-10W (N-series) CPUs will stick with Alder Lake since Arrow Lake is expensive to manufacture and package.
    I don't think expensive Arrow Lake is the reason they couldn't make a Skymont-based N-series. IIRC they made changes that make it easier to port a design between TSMC and Intel nodes (help me find the article if you can). They would be likely to make a smaller N-series die monolithic instead of a tiled design (if that is apparently more expensive).

    Maybe they don't feel the need to update a low-margin product which is easy to make on their older node.
    Reply
  • Notton
    If the power consumption was low and prices reasonable, I would love to see this in a >6 slot Asustor Flashtor. Their current 6/12 slot NAS' using N100 w/ PCIe switches leaves a lot to be desired.
    Reply
  • abufrejoval
    T-types have been a contant cause of frustration with me.

    Mostly I'd say because Intel's drive to constantly segment the market has then left many of those segments too small to be viable for OEMs and resulted in more and more niche plots going empty over time.

    One of the earliest 24x7 PC or µ-server appliances in my home was a pfSense firewall/proxy, which would augment the one in the router I just didn't trust enough to keep the garbage out. Especially since kids and wife weren't at all security conscious at the time. I started with an old notebook (single core Merom or Dotah, if I remember correctly) for the low-power, but once the Internet bandwidth Steam-ed up, something bigger was required.

    I got my first start on Atoms with a Bay Trail J1900, but that couldn't keep up once packets were inspected more deeply, so I concluded I really needed Core power, but without the consumption or noise of a full desktop PC.

    Nearly everbody was absolutely crazy for top clocks, while I wanted top performance for a given power and cooling envelope, ideally within the then notebook range of then 35 Watts tops. But nobody sold notebook chips in a Mini-ITX form factor at the time, only Atoms which were a little too weak.

    Only the T-variant of the then top-of-the-line i7-7700 series seemed to fit, which wasn't any cheaper than the i7-7700, only supported lower PL1 and PL2 settings: otherwise it seemed identical. So why wouldn't Intel simply allow lower TDP for normal Core CPUs to cover that?

    Anyhow, the ability to on one hand run CPUs at near normal desktop power and performance, yet also be able to curtail their Wattage for 24x7 µ-server use has been a constant pain over the last decade. I constantly rebuild system and like to move parts around so this loss of flexibility has cost me quite a bit over the years (exactly as planned by Intel, I guess).

    But of late mobile on desktop designs have become a thing, especially for the Chinese domestic market, I guess, but it's spilling globally via Aliexpress and others. And while notebook chips command a premium when they come out (more expensive than T-types), they fall out of fashion just as quickly and are sold at high bargains a generation or more later. Hence this mobile-on-desktop market from smaller Chinese OEMs.

    Case in point: I got myself a Minisforum BD790i the other day, a Zen 4 Ryzen 9 7945HX 16-core on a Mini-ITX board, which comes with a 100 Watts power limit by default (and overclock abilities), but can also be configured to 55, 45 and 35 Watts and runs amazingly cool with just a small fan while providing quite significant compute power.

    At €500 for the full 16-core system and 24 lanes of PCIe v5 it's quite an attractive alternative to what you'd have to spend on a Zen 5 8-core these days (and the same price as a 7950X CPU). It might not do games at 400 FPS on an RTX 4090 in TrueHD, but while it wouldn't do badly in any gamer setup, it's main attraction is that it's usable over such a vast range of applications, including a quiet µ-server that still packs quite a bit of punch for short peaks.

    Hyperscalers design for the optimal hardware at near constant >90% loads, my stuff tends to be often idle. But when there is things to be done, I want them done fast, just like on a notebook. And these new mobile-on-desktop systems offer just that, albeit at the cost of not having a socket, but often at a price where the mainboard is for free.

    And that's where I stop complaining about soldered SoCs.

    Now if Dragon Range only came with ECC support...
    Reply