News Intel Launches 14th-Gen Raptor Lake Refresh: Core i9-14900K, i7-14700K, and i5-14600K on October 17

Status
Not open for further replies.

bit_user

Titan
Ambassador
They've got the crosshairs on last gen Ryzen? Wow, Intel don't set the bar too high for yourself.
I think we don't know the precise timing on Arrow Lake's launch. It could happen close enough to Ryzen 8000 that they don't really need these CPUs to be competitive against it.

Remember, Intel originally planned to launch Meteor Lake on the desktop, this year. This Raptor Refresh is their plan B - seemingly a last-minute one, at that. That's why it's so underwhelming.
 
So... they took the most power-hungry consumer-grade CPUs ever made and overclocked them even further. What could go wrong?

I'm starting to think that Intel just might be in bed with Asetek.... ;)(y)
Nope, still 253W according to the article just like 13th gen and only 23W above the limit of the 7950x of 230W PPT.
You are talking about overclocking potential only here and it remains to be seen from reviews if the new ones are even better overclockers or not, but they are not higher (power) overclocked by intel.
 
  • Like
Reactions: Order 66

Order 66

Grand Moff
Apr 13, 2023
2,164
909
2,570
So... they took the most power-hungry consumer-grade CPUs ever made and overclocked them even further. What could go wrong?

I'm starting to think that Intel just might be in bed with Asetek.... ;)(y)
The CPUs are actually not any more power-hungry than last gen like TerryLaze said, but of course, you wouldn't have seen his post.
 

uwhusky1991

Reputable
May 7, 2020
17
10
4,515
  • Like
Reactions: Order 66

Giroro

Splendid
At some point during Microsoft's big Windows 11 push, Intel said their new processors need Windows 11 to run correctly. I think that was a mistake which continues to bite them.

Even though I know you can make Intel CPUs run just fine on Windows 10, my excitement for new processors is constantly being deflated by the threat that I might need to install Windows 11.
 
They've got the crosshairs on last gen Ryzen? Wow, Intel don't set the bar too high for yourself.
Well, to be fair, it is just a refresh of what they currently have. Their current architecture is already out-matched by Zen3 so expecting better from their current architecture would be overly-ambitious to the point of being delusional.
Still, no official Intel ARK page detailing specs.
It's probably still under embargo.
Pretty wild to have a new generation of chip's that still support's DDR4. That'll make a lot of older system owners happy.
I agree. I honestly believe that the adoption of DDR5 was a bit premature. Hell, I know people who are still rocking DDR3 and are doing just fine with it.
It was a joke, Asetek is a huge OEM of CPU liquid-cooled AIO solutions. Historically, a lot of manufacturers just bought some Asetek units in bulk and threw their names on them. I don't blame you for not getting it because it was a bit of an obscure reference and it's really my fault for not having linked the word "Asetek" for context. I've since fixed that oversight. ;)
The CPUs are actually not any more power-hungry than last gen like TerryLaze said, but of course, you wouldn't have seen his post.
You're right, I didn't see it. Hell, I don't even know who he is, but I wasn't being completely serious. I don't think that Intel would be crazy enough to do that.

(Although I also never expected Intel to one day produce a CPU even more power-hungry than the AMD FX-9590, which became a meme in itself.)
 
  • Like
Reactions: Order 66 and sherhi

bit_user

Titan
Ambassador
At some point during Microsoft's big Windows 11 push, Intel said their new processors need Windows 11 to run correctly.
Though it's probably not relevant to your needs, I would note that Linux has had support for the ThreadDirector for a few kernel revisions, now. I don't recall when the last big patch got merged, but it's safe to assume even Ubuntu 23.04 has support for it. Probably also 22.04 HWE, but worth checking for anyone who cares.
 
  • Like
Reactions: cyrusfox

Eximo

Titan
Ambassador
how so? (I don't like to accept statements automatically, I want to know the why not just that it happens)
I believe it is the 2x32bit addressing rather than the 1x64 bit addressing in DDR4 and earlier.

Sort of like quad channel, plus the extra bandwidth always helps with workstation type tasks.

As I recall, a lot of games show almost zero performance penalty with a single DDR5 stick. That will probably change over time.
 

HideOut

Distinguished
Dec 24, 2005
592
96
19,070
Their gaming benchmarks in the first slide show 1080p performance. Whos going to spend $500+ on a cpu and matching build for 1080p gaming? Show 1440p and above FFS.
 

bit_user

Titan
Ambassador
how so? (I don't like to accept statements automatically, I want to know the why not just that it happens)
A benchmark is worth 10k words, IMO.

Upon reviewing the benchmarks I can find for Raptor Lake, the results seem decidedly mixed. One of the cases to benefit more is this Google Chrome - Jetstream v2 benchmark, which showed a 7.3% advantage for DDR5-7200/34 vs. DDR4-3600/18.


VRay NEXT showed a 4.4% benefit:


In terms of gaming Watch Dogs: Legion showed a 5.8% benefit, at 1080p.


The above case were the biggest differences from that review, in benchmarks that weren't totally synthetic. They included numerous other cases where the difference was negligible.

However, the biggest improvement I'm aware of is in SPEC2017 rate-N. I believe the way this works is that N separate instances of each single-threaded benchmark are run simultaneously. This is a torture test for memory bandwidth, because there's no data-sharing between any of the cores/threads. Sadly, Anandtech only ran this test for Alder Lake, and didn't repeat it for Raptor Lake. However, it showed a whopping 31.3% benefit for integer and a 37.4% benefit for floating point.

117496.png

Source: https://www.anandtech.com/show/1704...hybrid-performance-brings-hybrid-complexity/8

SPEC2017 consists of 22 real-world applications. So, even though the way they run the rate-N test is a bit artificial, I don't consider it a totally synthetic benchmark.

So, in conclusion: I should amend my statement. It can make a big difference for heavily multi-threaded workloads, but it seems the difference is typically negligible vs. a high-performance DDR4 kit.

Still, if you're buying an i9 and want to squeeze the most performance from your system, it'd make sense to go with a fast DDR5 kit. It almost always helps, even if not by much. Furthermore, if you consider that the generational difference between Gen 13 and Gen 14 i9 processors is only about 2%, it puts some of the above gains in a different light.
 

bit_user

Titan
Ambassador
I believe it is the 2x32bit addressing rather than the 1x64 bit addressing in DDR4 and earlier.

Sort of like quad channel,
It's true that DDR5 subdivides a single 64-bit DIMM into two, independent 32-bit subchannels. This provides more parallelism in the hardware.

plus the extra bandwidth always helps with workstation type tasks.
Yes! Consider the sheer difference in bandwidth! The number following the DDR standard tells you how they compare. DDR5-7200 has exactly twice the theoretical bandwidth of DDR4-3600. Multithreaded workloads tend to be bandwidth-hungry, which is why the 96-core Genoa and 128-core Bergamo CPUs support 12-channel memory!! In the case of Genoa, that works out to 8x Zen 4 cores per channel, which is exactly the same as their top-end desktop CPU: the R9 7950X.

I think it's no coincidence that when AMD increased their top-spec server platform from 64 cores per CPU to 96, they also increased the max memory spec from 8-channel to 12-channel.

I even saw a recent interview with AMD's CTO (?), where the interviewer asked him if he foresaw > 16 cores in their desktop CPUs, and his answer was he didn't think it made sense in the near term, because a higher core count would be too bandwidth-starved.
 
Status
Not open for further replies.