Intel Core Ultra 9 285K Review: Arrow Lake is a Mess

[opckieran]

Intel always manages to beat AMD in the worst ways.

 

[hahahanoobs]

This is Intels first tiled chips. Basically the same as AMD Chiplets, only AMD had been doing this for close to a decade now.

They also ditched hyper-threading.

It's not a bad first attempt at tiled CPUs from Intel, but clearly far from perfect.

Newer Xeons, Ponte Vecchio, Meteor lake and Lunar lake are tiled parts.

 

[Kirby1]

Newer Xeons, Ponte Vecchio, Meteor lake and Lunar lake are tiled parts.

OK, first consumer CPU from Intel then.

 

[MarcusNumb]

The price must be a joke, another opportunity for AMD to gain market share if they price the 9800x3D right, and again cause it's AMD, I think next month we'll see the 9800x3d cost 500 xD Anw, I like this competition, unlike the GPU competition...

 

[maxxcool7421]

I've been reading reviews on this site for some time. They seem increasingly dubious, with observations and conclusions at odds with what other reviewers publish, based on a conception of what they think the product should be as compared to what it is and what it is, and then raging about the difference. Sometimes it seems like a childish anger at what it is technically and financially possible. I guess vitriol captures eyeballs but it also makes the value of the reviews kind of hinky. Which is too bad as these guys are clearly capable of technically good work.

How's that intel Check ?

 

[AdamNovagen]

Let's face It - We are reaching the limit of Moore's Law both in CPU and GPU (per $ at least), and that's a GOOD THING. We will be buying PCs for life, and devs will have to eventually optimize, rather than count on growing performance (per dollar & per Watt).

My dear, sweet, naive summer child... The day that they can no longer find new hardware to entice us and computers are expected to go years without upgrades, is the day that subscription models fully consume the hardware and software world and the components begin to fail and need replacing after only a few years' time. They always squeeze their revenue streams out somewhere... The line must go up.

Regarding this review: even as someone who has an entirely AMD-based PC at this point, I actually don't see this as a total loss for Intel. At the very least, I think this is a better showing than they've had for the last couple of generations; they've managed to put performance in the new chips that more-or-less matches the previous gen on average, while significantly reducing power consumption. While I don't think this generation will be worth buying, it makes me hopeful that they could at least be laying a good foundation here that could ramp up rapidly in the next couple of gens.

TO CLARIFY MY STATEMENT ABOUT THE PERFORMANCE: Yes, I saw the graphs, but frankly, a lot of what's going on here sounds like a very volatile software environment. Personally, I'm not actually counting Plague Tale, Homeworld or Cyberpunk in my opinion of this chip because they're such massive outliers that I genuinely think it's a software/driver optimization issue rather than an actual hardware problem. If you drop those three out of the comparison and factor in the productivity as well as the gaming, it basically looks like a slightly underclocked 14900K with lower power consumption (which, with a lowered boost clock, it essentially is). Don't get me wrong, I would not buy this actual chip, but I think it could at least be the start of something better.

 

[godrilla]

5% slower my arse. Hey at least they get to join the 10ghz ram kit club !

 

[m3tavision]

My dear, sweet, naive summer child... The day that they can no longer find new hardware to entice us and computers are expected to go years without upgrades, is the day that subscription models fully consume the hardware and software world and the components begin to fail and need replacing after only a few years' time. They always squeeze their revenue streams out somewhere... The line must go up.

Regarding this review: even as someone who has an entirely AMD-based PC at this point, I actually don't see this as a total loss for Intel. At the very least, I think this is a better showing than they've had for the last couple of generations; they've managed to put performance in the new chips that more-or-less matches the previous gen on average, while significantly reducing power consumption. While I don't think this generation will be worth buying, it makes me hopeful that they could at least be laying a good foundation here that could ramp up rapidly in the next couple of gens.

TO CLARIFY MY STATEMENT ABOUT THE PERFORMANCE: Yes, I saw the graphs, but frankly, a lot of what's going on here sounds like a very volatile software environment. Personally, I'm not actually counting Plague Tale, Homeworld or Cyberpunk in my opinion of this chip because they're such massive outliers that I genuinely think it's a software/driver optimization issue rather than an actual hardware problem. If you drop those three out of the comparison and factor in the productivity as well as the gaming, it basically looks like a slightly underclocked 14900K with lower power consumption (which, with a lowered boost clock, it essentially is). Don't get me wrong, I would not buy this actual chip, but I think it could at least be the start of something better.

Bruh...
They do not do "software/driver optimization" in games for CPUs. That is for GPUs...

A CPUs job is totally different and yes, there are OS optimization that can be had for CPU's (or even debugging), but that is NOT a game developer thing. No amount of anything is going to help this dumpster fire.


Coincidentally (as u mentioned) AM5 is coming up on 3 years now and anyone who (at the time) were shopping 12th Gen iNTEL are out of luck now. Arrow Lake should further teach a lesson to these people, knowing that iNTEL's LGA1700 was so short lived and NOW iNTEL is trying sell you another mobo and new CPU socket... in hopes of trying to equal AMD's almost 3 year old AM5 7800X3D.


One choice and it's real easy:
Buy an AM5 board and whatever CPU you think benefits you and enjoy. And a few years later or whenever you are feeling slow You'll be able to drop in a 10800X3D and never loose top-performance. EZ..

 

[Cracken]

Intel is trying to charge a premium for this release. So sad, this and the AMD 9000 series was a bust for both camps.

Well, maybe next generation.

9000X3D does look to be a pretty decent update, 500mhz higher base, 200mhz higher boost compared to 7800x3D

 

[PaladinNH]

We’re entering the monopoly stage here for AMD cpu’s and Nvidia gpu’s. Bad news for everyone - prices will skyrocket

High end gaming market, for sure. Hopefully AMD continues to produce the 7900 XTX. It's now the only alternative in this class. As for CPU, this release coupled with the Raptor Lake degeneration has knocked Intel clear out of the gaming CPU competition until Arrow Lake's next iteration. On paper, Intel seemed to have nailed the trifecta here. Single, multi, competitive efficiency. Shame it's so far behind in the real world.

 

[Fanboiism101]

I thought Arrow Lake is going to be a total lost cause across all benchmark, which make my expectation as low as it can be. But as reviews started coming out, Arrow Lake is not that bad. A mess, but not as bad as I initially thought, a far cry from what everyone is saying as they are able to beat the similarly prized AMD's in some productivity workload.

But, looking at their price and efficiency? I don't think AMD will lose much of their market right now.

A nice foundation, tho.

 

[Axeia]

I guess vitriol captures eyeballs but it also makes the value of the reviews kind of hinky. Which is too bad as these guys are clearly capable of technically good work.

Hardly vitriol, they just expected better.
They could have highlighted it a bit more that this was intels first go at multiple firsts. That the next generation might see a nice uplift (like AMD did going from Ryzen 1000 to 2000). But it wouldn't change the conclusion, this generation isn't worth buying. If they want a more positive review they should have launched at lower prices which is the whole reason to go for this times approach, saving costs.

 

[scoffer]

Very disappointing, by the look of things. Specially for the buying public. As mentioned elsewhere- just watch AMD and associated sellers start with the scalping prices. I still don't see why so many people see AMD as the underdog 'good guys'. What utter BS. As with any business -given the opportunity-they will scalp you to within an inch of your life.

 

[BadThad]

New CPU architecture, new platform, drivers and BIOS'. It will take some time to iron everything out. This isn't a horrible showing for Intel. Things will get better as it matures.

 

[Strawman]

Gaming performance not withstanding, it looks like an incredible CPU. MT performance and efficiency is on par with the 9950x (loses some, wins some), gaming performance is pretty much identical while it uses a lot less power on average (35-40w according to TPU and tom's hardware) and it's way more efficient on lighter load / st / idle.

Until the 9950x 3d releases, this is the GOAT. If the mediocre gaming performance gets fixed somehow (if it's a scheduler issue) it looks like a damn winner.

 

[Strawman]

I thought Arrow Lake is going to be a total lost cause across all benchmark, which make my expectation as low as it can be. But as reviews started coming out, Arrow Lake is not that bad. A mess, but not as bad as I initially thought, a far cry from what everyone is saying as they are able to beat the similarly prized AMD's in some productivity workload.

But, looking at their price and efficiency? I don't think AMD will lose much of their market right now.

A nice foundation, tho.

Overall across the 4 reviews I checked that do some more thorough power measuring all the intel chips seem to be more efficient than their amd counterparts. The big downside is the gaming performance, but even there the i9 is matching the 9950x. It ain't horrible, but no great either.

 

[YudaHnK]

Great review - it seems only a diehard Intel-fan would buy this flop.
I regret not buying 7800x3D when it was cheap this summer, the price will not go back to those levels now with Intel falling over themselves.

 

[Strawman]

Well what hub isnt telling you cause they only tested power draw in 2 games, the 9950x draws 40% more power in gaming for the exact same performance as the 285k.

Toms hardware and TPU have done a per game and an average gaming power draw and it looks like the 285k is just using way less power than the 9950x.

But sure, let's ignore that

 

[Lounds]

I'm glad I decided to buy a Ryzen 7 5700X3D for my old B350 motherboard instead of investing in a whole new platform. Best £161 I've ever spent. If I last 3 years that's 10 years on the same motherboard.

 

[fadingfool]

Well what hub isnt telling you cause they only tested power draw in 2 games, the 9950x draws 40% more power in gaming for the exact same performance as the 285k.

Toms hardware and TPU have done a per game and an average gaming power draw and it looks like the 285k is just using way less power than the 9950x.

But sure, let's ignore that

On average the 285k competes with the 7700X - how does that power comparison look now?

 

[Strawman]

On average the 285k competes with the 7700X - how does that power comparison look now?

You mean in gaming? It's literally competing with the 9950x. There are within 2% in most reviews, either of them is on top.

Sadly tom's hardware doesn't have the 7700x but it has t he 9600x and the 9700x. It's a lot more efficient than both of them in gaming.

 

[Eldrach]

Gaming performance not withstanding, it looks like an incredible CPU. MT performance and efficiency is on par with the 9950x (loses some, wins some), gaming performance is pretty much identical while it uses a lot less power on average (35-40w according to TPU and tom's hardware) and it's way more efficient on lighter load / st / idle.

Until the 9950x 3d releases, this is the GOAT. If the mediocre gaming performance gets fixed somehow (if it's a scheduler issue) it looks like a damn winner.

If they, as rumored - will share the X3D cache between the chiplets + increases the size of the X3D Cache for the 9950, then we’re probably looking at the best gaming and productivity cpu combined - as long as they get the scheduling done correctly

 

[hahahanoobs]

OK, first consumer CPU from Intel then.

MTL and LNL are consumer parts.
If you mean retail boxed consumer parts, then yes, ARL is the first.

 

[Theinsanegamer]

Well what hub isnt telling you cause they only tested power draw in 2 games, the 9950x draws 40% more power in gaming for the exact same performance as the 285k.

Toms hardware and TPU have done a per game and an average gaming power draw and it looks like the 285k is just using way less power than the 9950x.

But sure, let's ignore that

TPU shows the 9950x scoring 0.1 points HIGHER than the 285k in efficiency on productivity benchmarks.

In gaming, the 9950x scores 1.85 FPS/w, while the 285k is 2.01. That's not a 40% difference. This also ignores the now 3 year old 5800x3d which not only beat intel in gaming but also scored 3.05 FPS/2, or the 7800x3d which hit 4.25.

You mean in gaming? It's literally competing with the 9950x. There are within 2% in most reviews, either of them is on top.

Sadly tom's hardware doesn't have the 7700x but it has t he 9600x and the 9700x. It's a lot more efficient than both of them in gaming.

How are they measuring? Toms hardware reports 77.2 watts in baulders gate 3 for the 285k. Gamer nexus shows the same CPU, same game, at 87.9 OR 112.5, depending on how thoroughly you want to measure. Starfield, toms reports 140w, GN 172.

Intel is drawing significant power from the 24 pin, unlike the 14th gen or earlier chips, and sites like toms (who lost all credibility during the "just buy nvidia" scandal years ago) do not report how they measure power from these chips. As such I find it highly unlikely their efficiency data is correct.

 

[Strawman]

TPU shows the 9950x scoring 0.1 points HIGHER than the 285k in efficiency on productivity benchmarks.

So they are identical in MT efficiency, just as I've said.

In gaming, the 9950x scores 1.85 FPS/w, while the 285k is 2.01. That's not a 40% difference. This also ignores the now 3 year old 5800x3d which not only beat intel in gaming but also scored 3.05 FPS/2, or the 7800x3d which hit 4.25.

How are they measuring? Toms hardware reports 77.2 watts in baulders gate 3 for the 285k. Gamer nexus shows the same CPU, same game, at 87.9 OR 112.5, depending on how thoroughly you want to measure. Starfield, toms reports 140w, GN 172.

Intel is drawing significant power from the 24 pin, unlike the 14th gen or earlier chips, and sites like toms (who lost all credibility during the "just buy nvidia" scandal years ago) do not report how they measure power from these chips. As such I find it highly unlikely their efficiency data is correct.

Both igors lab, TPU and Tom's hardware had the 285k efficient in gaming than the 9950x. The 40% is taken from Tom's hardware (76w vs 111w on the 14 game average). It doesn't really matter if you accept the 40% difference from Tom's, cause it's more efficient on every review anyways.

No, intel is not drawing significant power from the 24 pin. You are flat out lying or confused yourself. That's only the asus mobos doing it, it has nothing to do with the cpu, asus mingled some cpu vrms into the atx 24pin. TPU used an msi mobo to measure power for specifically that reason and lo and behold, 285k was still more efficient in gaming.

I'm not ignoringthe 5800x 3d and the 7800x 3d but those are megaslow cpus on every other task, they are not meant to compete with the 285k. The 9950x is. Nobody looking to get a 9950x is remotely interested in the 5800x 3d for obvious reasons.