Intel Core Ultra 9 285K, Ultra 7 265K and Ultra 5 245K Review Roundup

DPI said:
It's just a bummer that after 9000X3D releases it'll be a deadzone until 2026/2027, since both companies have deprioritized consumer dies because they can't count the money fast enough on the Enterprise side.
Click to expand...
What the hell is driving cpu demand on datacenter market?
 
atp1916 said:
Well if this Intel launch wasn't enough to keep me AMD i don't know what else could be.

Still running an AM4 5950x build (it's not terrible) - but...

I GOT AN ITCH.
Click to expand...
"The only way to get rid of a temptation is to yield to it. Resist it, and your soul grows sick with longing for the things it has forbidden to itself, with desire for what its monstrous laws have made monstrous and unlawful." O.W.
 
So it's slower in gaming than normal Zen 5 let alone the X3D parts. IMC going off die sure hit them hard, and AMD manages to be ahead of them in gaming even with their archaic IMC.

Disappointing, I was hoping for a good release with price cuts from both camps. Instead, 9950X goes off stock pretty much everywhere overnight.
 
1.1.2.3.5... said:
What the hell is driving cpu demand on datacenter market?
Click to expand...

Enterprise has always been a bigger profit center per unit area of wafer. No change there. Consumer only ever makes sense when it is a volume business.

Essentially, both on the GPU and CPU side, enterprise usually pays for the development of a new arch, and consumer gets the hand me downs. Once the arch has already been developed for enterprise, you can pad the margins by also selling derivatives or lower bins to the consumer market.

It's been a long time since consumer CPU development paid for itself.
 
Last edited:
If this isn’t just a microcode issue things aren’t going to be looking good for Intel. This release was already pretty lacklustre but I figured Intel still end up selling lots in prebuilt systems to people who buy based on the Intel sticker. If this is bad enough that the Dells of the world don’t want the hassle it’s going to be DOA.

Pretty shocking really. I can remember when Intel was not only the fastest but the most most stable hardware you could buy, if you were building a system that’s what you wanted, not some Cyrix or AMD CPU. Since then AMD has done amazing things for certain but Intel
Has faltered and this could be more of a stumble or the beginning of a fall. Who knows? It’s certainly interesting to watch.

Then again maybe it’s a microcode fix away from something more impressive? Maybe it needs a voltage (and resulting power use/thermal) bump for stability? Maybe Intel sandbagged so they get a huge amount of coverage after pulling out a microcode home run blowing away the competition??

Can’t wait for the next dramatic turn LOL

I have run AMD/Cyrix/Intel in the past. Currently on AMD and perfectly satisfied. Can’t see switching any time soon at this rate. Hate to see less competition though, it’s not good for pricing!
 
gtrguy said:
I have run AMD/Cyrix/Intel in the past. Currently on AMD and perfectly satisfied. Can’t see switching any time soon at this rate. Hate to see less competition though, it’s not good for pricing!
Click to expand...
Intel has been overly dominate in terms over overall sales and I think they still have the most. Giving market share to AMD I think will help prices as Intel will have to reduce margins and become a leaner company.
 
Last edited:
gtrguy said:
If this isn’t just a microcode issue things aren’t going to be looking good for Intel. This release was already pretty lacklustre but I figured Intel still end up selling lots in prebuilt systems to people who buy based on the Intel sticker. If this is bad enough that the Dells of the world don’t want the hassle it’s going to be DOA.

Pretty shocking really. I can remember when Intel was not only the fastest but the most most stable hardware you could buy, if you were building a system that’s what you wanted, not some Cyrix or AMD CPU. Since then AMD has done amazing things for certain but Intel
Has faltered and this could be more of a stumble or the beginning of a fall. Who knows? It’s certainly interesting to watch.

Then again maybe it’s a microcode fix away from something more impressive? Maybe it needs a voltage (and resulting power use/thermal) bump for stability? Maybe Intel sandbagged so they get a huge amount of coverage after pulling out a microcode home run blowing away the competition??

Can’t wait for the next dramatic turn LOL

I have run AMD/Cyrix/Intel in the past. Currently on AMD and perfectly satisfied. Can’t see switching any time soon at this rate. Hate to see less competition though, it’s not good for pricing!
Click to expand...
I would argue ntels fall began all the way back in 2001 back when Pat was CTO. It takes a long time to screw up a company like Intel. IMO it started when their culture started to sniff their own expulsion. Before that time Intel was a confident company but I believe enough of the early people were still there that they had had some humility. Around the time they designed Sandy they were starting to believe no one could touch them, that they could solve any engineering problem easily because they were Intel. Pat the current CEO was part of that gen of Intel will win because Intel is Intel management. He set a path for Intel foundry that was far to aggressive, because we are Intel and all the issues people point out we will crush, we are Intel after all. When Pat took over as CEO he spoke the same way. We will Intel because Intel Intels.

I have said it a bunch of times. Intel hasn't been as far ahead in performance as they were given credit for even prior to AMD Zen. Intel started making a lot of odd design choices going back a long ways now. A combination of Luck and paying security researchers to shut up saved some of their big hits. Take as an example Sandy Bridge. Fondly remembered as the gen that laid the boots to AMD and forced AMD to throw out their designs and start on Zen. (which was for the best no doubt). Imagine if the massive security holes that frankly Intel knew about from its design phase was published in 2012 and not 2018. [You can't not know your chip is completely skipping checking permissions on cache memory at all] The first paper to detail those attacks was published in 1995! "The Intel 80x86 Processor Architecture: Pitfalls for Secure Systems" Looking back its a how to manual to perform side channel memory attacks on x86. Apple mitigated for these exploits in the MacOS kernel in 2012. Researchers found Specter and Meltdown attacks in '16 and were convinced to stay quite on them for 2 years.

This is a modern retest of Sandy and FX. Granted Sandy is faster even with mitigations. No doubt. Just imagine though if like the 13th and 14th gen where the burn out issue happened while they were still on the shelf. Imagine if everyone rocking their year or two old Sandy bridge got security updates that dropped their performance 30-40%. I mean Sandy bridge vs FX... it was 40% faster when they were on the shelves one at $500 one at $200. Had Intel been forced to mitigate for their terrible cache design while they were current that would have dropped to 15% faster at 2.5x the cost. Intels luck has been running out the last year or so... pushing insane power through their chips to keep up has bit them in the ass. It looks like having to address their issues with power use, and failing fab buinsess has resulted in what ever this ultra gen is going to turn out to be. Maybe its all software and they can fix it? Maybe.

432aff7.jpg
 
Well if anything I won't go AMD until Zen 6 comes out it would feel like a downgrade to have a CPU without ecores.
 
Dan_D said:
I remember doing the testing on this at the time. The difference between Zen 1/2 and Intel CPU's in games was a lot bigger than that............at 1920x1080. The argument was that the Zen CPU's were often better at everything else once their core counts went up and of course, when you were in a more GPU limited scenario such as 4K gaming. Even Zen 3 didn't achieve parity with Intel in gaming though it was a lot closer.

Now I do remember a lot of sites showing similar average FPS numbers, etc. on Zen 2 CPU's and even Threadripper. But I had a 2920X and it was garbage even at 4K gaming. When I dug into that, I realized that Zen was giving me much higher maximum frames but also much worse minimum frames than the 9900K did at the time. The average reports would have led you to conclude parity, but in the actual gaming experience this wasn't the case. This is why I went back to Intel despite otherwise really liking my Threadripper 2920X.
Click to expand...
Yup. Even my 3960x has things where you can feel micro stutters compared to my 10900k - they were designed for different things. My 3950X is running 4k gaming when not a server, and same dance there - Intel of that gen was better at gaming, although Zen was very good at it... but Zen was excellent at being a high end workstation, while Intel was just ok. Hence why I built two.
 
mkppo said:
So it's slower in gaming than normal Zen 5 let alone the X3D parts. IMC going off die sure hit them hard, and AMD manages to be ahead of them in gaming even with their archaic IMC.

Disappointing, I was hoping for a good release with price cuts from both camps. Instead, 9950X goes off stock pretty much everywhere overnight.
Click to expand...
This isn't the first time Intel moved the memory controller off die. The first time I recall them doing it was with first generation Core (Clarkdale). I don't think they moved the IMC back onto the processor die until 11th generation Core (Raptor Lake).
 
Dan_D said:
There are certain advantages to changing sockets more frequently. You inherit a lot of legacy baggage building CPU's generation after generation to be compatible with motherboards upwards of half a decade old.
Click to expand...
In this generation at least it makes sense compared to older ones since 12/13/14 supported both DDR4 and DDR5 but they removed the DDR4 controller for Core Ultra. If they supported it on older sockets you know some genius would try sticking one in a DDR4 board lol.
 
kac77 said:
MLD actually did a flashback to Zen 1 within his Ultra 9 285 review. It's cool to see how performance was back then. I forgot that Ryzen's lows in gaming were often better this was why there was such a nuanced view of it.
View attachment 687569View attachment 687570
Click to expand...

BF1 is a best-case scenario for Zen1. I 'upgraded' from an i5-6600k to a Ryzen 7 1700x and had a massive loss of performance in many games. I don't remember all of the specifics, but I remember that League of Legends dropped from a solid 130fps on the 6600k to a fluctuating 80-90fps on the 1700x. Most games at the time were about 15% worse on Zen (max-fps) as compared to the 6600k IIRC. I ended up returning it for a 7700k which wasn't a big upgrade max-fps wise over the 6600k but it had some measurable min-fps gains in some games while also being the last hurrah for that platform. Now, I do believe that the issues with Zen1/2 were related to it being such a new architecture and suffering from a combination of early BIOS issues, bad/slow memory support, the Windows scheduler not properly supporting them, AND the fact that most games hadn't yet made the move to utilizing more than four cores. But still...

That 6600k is still trucking along in mom's PC and she thinks it's super fast.
 
  • Like
Reactions: cycy
like this
Lol. LTT rating their gaming graphs in rank by 1% lows. Who does that? I guess it was the only way to not completely let Intel look like a slouch.


I think this may be Intel's Zen1 moment. Refresh on this be a major improvement I think. Isn't this also the first part out under Jim Keller?
 
alxlwson said:
Lol. LTT rating their gaming graphs in rank by 1% lows. Who does that? I guess it was the only way to not completely let Intel look like a slouch.


I think this may be Intel's Zen1 moment. Refresh on this be a major improvement I think. Isn't this also the first part out under Jim Keller?
Click to expand...
Jim left in 2020. They cancelled his royal core project. Pat shit canned all his work.
 
alxlwson said:
Lol. LTT rating their gaming graphs in rank by 1% lows. Who does that? I guess it was the only way to not completely let Intel look like a slouch.


I think this may be Intel's Zen1 moment. Refresh on this be a major improvement I think. Isn't this also the first part out under Jim Keller?
Click to expand...

Eh, I take 1% lows more seriously than max fps most of the time but to rate hardware solely by 1% lows is silly.
 
  • Like
Reactions: Dan_D
like this
alxlwson said:
I think this may be Intel's Zen1 moment. Refresh on this be a major improvement I think. Isn't this also the first part out under Jim Keller?
Click to expand...
Obviously, we can't know for sure at this point but a lot of Intel's more transitional products paid off in later iterations.
 
It could but this is already a refresh of Meteor lake (cancelled on desktop) I think ? And rumours seem to be to revert a lot of what is left for the next one...

If the rumours is true, it seem that it get less and less ambitious on the tile concept (and the money saving they could do). Could be all wrong, but I feel like Meteor lake was the platform launch, and this launch was supposed to be the working refined iteration of it, does not feel like it because of how little noise ML ended up doing (and no desktop version launched obviously, adding to the lack of noise).
 
Last edited:
Okatis said:
Damn, these results are brutal. Game framerates average the same as the i7-12700K, with some results worse than even the i5-12600K.
Click to expand...

The 12700k is so cheap now at $219 in the US that it is the far better deal than the 245k.
 
Armenius said:
This isn't the first time Intel moved the memory controller off die. The first time I recall them doing it was with first generation Core (Clarkdale). I don't think they moved the IMC back onto the processor die until 11th generation Core (Raptor Lake).
Click to expand...
Wasn't Haswell-e and Broadwell-e on die IMC? I remember my old 5960X having insanely low memory latency... even being the first Quad DDR4 platform.
 
The sad part of all of this is we (gamers) are going to be put further and further back in priority. We are low volume and low margin. Lunar Lake is arguably if not actually a win. Dell sells 8 million (don't know) business laptops a year and they refuse to put AMD in their Latitude series. In the end, we are a pittance of volume compared to the laptop market which is high volume low margin. The datacenter though, is an oddity because it has a lot of volume and good margin, AMD is killing Intel in that segment.
 
You must log in or register to reply here.
Back
Top