Question Raptor Lake - Official Thread

[Hulk]

Since we already have the first Raptor Lake leak I'm thinking it should have it's own thread.
What do we know so far?
From Anandtech's Intel Process Roadmap articles from July:

Built on Intel 7 with upgraded FinFET
10-15% PPW (performance-per-watt)
Last non-tiled consumer CPU as Meteor Lake will be tiled

I'm guessing this will be a minor update to ADL with just a few microarchitecture changes to the cores. The larger change will be the new process refinement allowing 8+16 at the top of the stack.

Will it work with current z690 motherboards? If yes then that could be a major selling point for people to move to ADL rather than wait.

 

[H433x0n]

You lost me on the "Eco" mode. Do you mean a setting through XTU? I mentioned it a while back but I would prefer if Intel rebuilt XTU to look more like Ryzen Master. RM is more easy to follow visually than reading various settings in XTU. Over the years XTU has not evolved from looking like a trading terminal monitor.

I mean as an option in the BIOS, similar to how Asus offers multicore enhancement. Granted, I personally don't care much about the power consumption since electricity is cheap where I'm at but it'd be a nice selling point.

That's basically the main complaint about Raptor Lake. If it suddenly used 20% less power and topped out at 200 watts, I don't see how it doesn't immediately become really compelling buy for anybody looking at a CPU (I already think it is, but I might not be the average consumer).

 

[Henry swagger]

Incorrect. L2 cache is getting an increase among other changes Not to take from the Intel thread but AMD's design with the extend cache layer is L3 cache. L2 is faster. My best guess going forward is AMD looked back in how great the 5775c was and expanded on that concept. AMD's approach is placed on top of the chiplet by vias. It's a modified approach vs a ground up approach. Culminating with other changes Raptor's refresh will put it slightly ahead of Zen 4 3d.

The real battle is next gen next year. All of Intel's new tech and core designs plurality included vs whatever Zen 5 is ignoring all the bs mlid or that fat kid with the mop hair post on youtubes.

If Intel is smart they'll come in at a lower price than the 3D with some incentives. June tru August is their launch date, and Zen 5 may very well launch between June and July of next year. Plenty of time for Intel to grab up sales. Both companies current hardware is amazing and competent. Power usage ignored if it doesn't affect you too much financially you can't go wrong with either direction.

My own hesitance with Zen 4 is knowing it may very well be a waste of money based on the claism I've heard of Zen 5, and I know through gut feeling X770 is going to be the board to have. PCIe 6 availability by Intel and AMD is likely. Depending on how the economy is in 1 years to 2, we may see as speedy as a rampup of DDR6 in 2026 as we did with DDR5.

Consumer hardware hasn't been this exciting in nearly 20 years. IDK how to embed a video but someone embed a video of people dancing or pretend it's there.


edit: i should note I expect intel to increase caches in their native ip without relying on the same method amd does. it adds risk and complexity to designs but it will avoid the minimal latency available on v-cache. it will also help with future thermals and be able to push frequencies higher than amd can. just gotta find that one sucker and convince them to buy an arrow lake or panther lake system and watch how good or awful the bugginess is.

I wont lie v-cache by amd is brilliant because of the power efficiency it gets and performance over standard zen 4 and 13900k.. intel needs to counter it if zen 5 has the rumoured 30+ ipc jim keller said

 

[A///]

I wont lie v-cache by amd is brilliant because of the power efficiency it gets and performance over standard zen 4 and 13900k.. intel needs to counter it if zen 5 has the rumoured 30+ ipc jim keller said

It is but as i pointed out it has its flaws. there is a minor latency. Voltages, frequency and heat soak are all issues with laying cache on top of the ccd chiplet. it is the simplest method for amd so they don't have to design a new ccd to include larger caches. they save a ton of money they wouldn't otherwise want to spend. For intel they have the option of incorporating big cache at L2 and L3 if they wanted to. I have to go back and watch raja 🤢 explain the design choices intel made a few years ago for future designs but i don't recall separate cache from the core and cache complexes. Intel can avoid some of the problems amd has such as having to run lower volts/being volt sensitive and thermal issues. this is why raptor lake refresh is so interesting as a first step. assuming assumptions are true and it's raptor lake with fatter cache, higher frequencies and using less power than plain jane raptor. If they knock it out of the field they're good until next June. The chances of amd spinning up xt chips isn't likely.

yes jim keller and mike clark have been blowing smoke up everyone's rear ends with Zen 5 for a few years now. Talk is cheap, they need to deliver.

 

[A///]

I mean as an option in the BIOS, similar to how Asus offers multicore enhancement. Granted, I personally don't care much about the power consumption since electricity is cheap where I'm at but it'd be a nice selling point.

That's basically the main complaint about Raptor Lake. If it suddenly used 20% less power and topped out at 200 watts, I don't see how it doesn't immediately become really compelling buy for anybody looking at a CPU (I already think it is, but I might not be the average consumer).

That's what I've been saying. Until AMD's recent price cuts the lower end processors were not compelling enough to buy over Intel, and now you can get a 7950x in a combo deal online for cheap.

For me as I told Mark this year I estimate the summer season to cost me an extra $100 based on how the pricing structure and how much I use. The amd's running "cooler" and lower power will cost me less. To replace all my Intels with AMDs will save me several hundred dollars in power bills a month due to them being left to their own boost levels.

 

[igor_kavinski]

To replace all my Intels with AMDs will save me several hundred dollars in power bills a month due to them being left to their own boost levels.

Get one 7950X, one 7950X3D and downclock the other remaining Intel's into their power sipping voltage zones and you should still save a lot. Then you can replace those Intel's next year with Zen 5 SKUs.

 

[DAPUNISHER]

Get one 7950X, one 7950X3D and downclock the other remaining Intel's into their power sipping voltage zones and you should still save a lot. Then you can replace those Intel's next year with Zen 5 SKUs.

I personally appreciate that you offer helpful advice to other members so freely. That said, please stop pimping AMD in an Intel thread. Shoot him a PM or something.

 

[BoomerD]

I'm having MINOR heat issues with my i5-13600K and Arctic Cooling Liquid Freezer II 280mm variant.
For the most part, the cooler does a great job keeping temps around 40C...except under loads. Then, cores 3 and 5 seem to get the hottest.


With the IC Graphite thermal pad and the radiator top-mounted in the be quiet! Dark Base Pro 900 Rev.2 case:

With Arctic Cooling's MX-5 TIM and the radiator mounted in the front of the case:

I'm going to try to re-do the IC thermal pad since the difference is pretty drastic between the TIM and pad, but my biggest concern is cores 3 and 5.

Anyone have a "map" that will tell me where those particular cores are in the CPU? I've looked and googled...find nothing useful.

I DO know that the paste application that I thought was poor...was actually pretty decent:

Thanks to a screwup by Amazon, I now have TWO tubes of Arctic's MX-6 TIM and cleaning pads. If I can't figure this out with the graphite pad, I'll try that next...but (correct me if I'm wrong...as usual) is it possible that despite what looks like good paste spread, I didn't get the two problematic cores "covered" well, leading to the heat in the previous mount?

 

[igor_kavinski]

How about just undervolting cores 3 and 5 in BIOS if that option is exposed for you (I recall that Intel left that up to mobo manufacturer to decide)?

 

[Mopetar]

It's not unusual for some cores to get hotter than others. The two that get the hottest are just physically located in an area of the chip that's closer to other active transistors so they can't shed heat as quickly.

I've got the same on my CPU where some cores run cooler even when they're all fully loaded, just because they're in a less active part of the chip and have dark silicon to dump some of the heat into.

You can get better cooling solution that you're pulling more heat off the CPU, but I'm willing to be the cores sandwiched between two other cores on either side are the ones that are hotter. Or it might be the ones near the E-cores, but the ones on the edges will be cooler no matter how good your cooling is.

 

[A///]

Remove your mx5 and order some mx6. mx5 got nixed due to counterfeits and poor formulation when it wasn't counterfeit. mx6 solved those issues. Try another paste brand.

hot pump cores are normal in this day and age. your cooler should be working fine because those arctics are bricks. have you tried adjusting your fan curves?

 

[BoomerD]

Remove your mx5 and order some mx6. mx5 got nixed due to counterfeits and poor formulation when it wasn't counterfeit. mx6 solved those issues. Try another paste brand.

hot pump cores are normal in this day and age. your cooler should be working fine because those arctics are bricks. have you tried adjusting your fan curves?

The second pic is Arctic MX-5 and the 280mm radiator mounted in the front of the case with the fans drawing outside air.
The first pic is with the IC Graphite pad and the radiator mounted in the top of the case and the fans drawing air from inside the case.
Like I said, due to a screwup by Amazon, I have two tubes of MX-6 and the Arctic cleaning pads. That might be the next thing I try. the graphite pads are new to me...and were highly recommended here.

 

[A///]

The second pic is Arctic MX-5 and the 280mm radiator mounted in the front of the case with the fans drawing outside air.
The first pic is with the IC Graphite pad and the radiator mounted in the top of the case and the fans drawing air from inside the case.
Like I said, due to a screwup by Amazon, I have two tubes of MX-6 and the Arctic cleaning pads. That might be the next thing I try. the graphite pads are new to me...and were highly recommended here.

Graphite pads are new to normal consumers but they've been in use in industrial for a while longer. Try the mx6. first and see how you like it. I had not read your post further than mx5 my apologies if I missed what you said. If I see mx5 my first thought is to reply about mx6. there are a lot of fine thermal compounds that aren't reactive nowadays.

 

[BoomerD]

Graphite pads are new to normal consumers but they've been in use in industrial for a while longer. Try the mx6. first and see how you like it. I had not read your post further than mx5 my apologies if I missed what you said. If I see mx5 my first thought is to reply about mx6. there are a lot of fine thermal compounds that aren't reactive nowadays.

No problem. Arctic included a dab of MX-5 with the cooler. I also have a partial tube of Noctua's NT-H1 TIM...but it's about 12 years old...so it's probably not so good anymore.

 

[Abwx]

I DO know that the paste application that I thought was poor...was actually pretty decent:

Thanks to a screwup by Amazon, I now have TWO tubes of Arctic's MX-6 TIM and cleaning pads. If I can't figure this out with the graphite pad, I'll try that next...but (correct me if I'm wrong...as usual) is it possible that despite what looks like good paste spread, I didn't get the two problematic cores "covered" well, leading to the heat in the previous mount?

In principle heat transfer from every part of the die will be equalized by the SKU metal pad, the difficulty is to have perfect homogeneity of the thermal contact between the SKU case and the heatspreader.

Good quality paste will often compensate a little for poor homogeneity but at the end of the day it will be inferior to an average paste that is very well applied, wich mean just enough paste to fill all surface micro irregularities but not too much because then it will also increase the case/heatspreaser thermal resistance.

This wasnt a problem with previous CPUs but with current 200W+ products any miss is
paid cash with abnormaly high temps for the less favoured cores.

 

[Timur Born]

12 E cores use the same space as 4 P cores:

7Z benchmark - 4 threads on 4 non HT P cores (5.6 Ghz):

7Z benchmark - 8 threads on 4 HT P cores (5.6 Ghz):

7Z benchmark - 12 threads on 12 E cores (4.3 Ghz):

 

[Timur Born]

 

[A///]

12 E cores use the same space as 4 P cores:

4 cores to a p core. but 2 e cores to the performance of 1 p core?

 

[Timur Born]

4 cores to a p core. but 2 e cores to the performance of 1 p core?

12 E cores = 4 P cores in space (a little less actually)
3 E cores = 1 P core in space

The pure E cores are smaller, but the shared L2 cache and bus per cluster also need space. See the last image I posted.

 

[Timur Born]

Intel i9 13900K 6GHz Thermal Velocity Boost Overclock Guide | MSI Z690 Unify - YouTube

That overclocked (in video) 13900K vs. my undervolted one:
- 6.6% higher CB23 multi-core score for 38% higher CPU Package Power.
- 6.7-7% higher CPU-Z multi-core score for 36% higher CPU Package Power.
- 3.3% higher CPU-Z single-core score for 35% higher CPU Package Power.

 

[A///]

12 E cores = 4 P cores in space (a little less actually)
3 E cores = 1 P core in space

The pure E cores are smaller, but the shared L2 cache and bus per cluster also need space. See the last image I posted.

no this I understand I was talking about performance in the 2nd half. unless that is what you meant.

 

[A///]

- 6.6% higher CB23 multi-core score for 38% higher CPU Package Power.

as I was saying about intel needing to trim the fat off their power use.

 

[Timur Born]

no this I understand I was talking about performance in the 2nd half. unless that is what you meant.

E cores are good at Integer workload like 7Z. An E core running a single thread of 7Z at 43x performs the same as a P core at 40x, both also use the same power then.
E core fall flat with floating-point load, but still better than having no extra cores for that.

 

[Timur Born]

as I was saying about intel needing to trim the fat off their power use.

Keep in mind, though, that overclockers drown their CPUs in voltage and water. That Youtuber is using 1.37V fixed Vcore coupled with some LLC increase (impedance decrease).

 

[A///]

E cores are good at Integer workload like 7Z. An E core running a single thread of 7Z at 43x performs the same as a P core at 40x, both also use the same power then.
E core fall flat with floating-point load, but still better than having no extra cores for that.

that is not what I asked. give me the basic performance. are 2 ecores the same as 1 p core in performance or not? giving me a breakdown of specific use cases doesn't help me when I don't use 7z or any program that uses similar logic.

 

[Timur Born]

No.