INTEL CPUs *OC Benchmark & Discussion*

[matique]

In the 3 generations, intel created an unnecessary problem for themselves due to political reasons.

Their driving force behind this hybrid architecture is likely due to their already subpar power efficiency at stock. Without those 'efficiency' cores, what's already bad would have been worse.
y
They are trying very hard to give the impression that they are still competitive, but anyone with a functioning brain can easily see through this.

Lastly, it makes no sense for other developers or vendors to work with them for their political agenda. Their hybrid architecture is FAR from something I would consider a technological advancement, it's literally just a cheapskate method that has known side effects. Rn, those vendors are just playing along out of respect.

Intel should stop their nonsense, stop expecting others to play along with them, be upfront, and rid of it altogether. Even if they can't keep up with the competition, they shouldn't be introducing such unnecessary issues. They should stop trying to get other vendors or companies involved in it.

Why must other companies waste time and resources to work with their nonsense? .

E-cores make sense when it's deployed well, such as handling background tasks and keeping idle power consumption low, i.e. 7-10w idle on intel vs 30-40w idle on AMD. Ecores do help out quite a bit with multicore processing, that can't be denied. 12th gen was a massive improvement over 11th, and 13th gen had considerable improvements as well with regards to clock speed, ipc as well as higher ring clocks not being tied down when ecores are enabled. 14th gen is just a joke.

Where intel messed up is the integration of their core deployment with windows especially when 12th gen first launched. Whose to blame, not sure on that. Windows? Game devs? Intel themselves? But currently it's pretty much fixed in most situations, especially with latest windows updates and when programs are updated accordingly. They've also messed up with APO, where it's limiting it only to its halo products despite there being very very little hardware differences between the generations.

Not even sure where you're coming from with regards to political agenda? FYI AMD is also moving to P and E core tech, if anything, working with intel provided with valuable lessons learnt for vendors to avoid the same issues with AMD.

With regards to keeping up with competition, head to head, 13th gen outclassed non x3d ryzen 7 series for each class. Props to AMD, their x3d tech is amazing especially for gaming. For those interested in "competition", i.e. enthusiasts seeking the fastest, 13900k/14900k is still the leading CPU when tuned up, whereby the ceiling is much higher for both CPU and RAM. You can't exactly tune an x3d chip since max freq is locked, and RAM on AM5 is throttled due to IF limitations. A tuned intel rig will not be as limited as a x3d chip when a game isn't able to be cached effectively. When L3 cache runs out, there will be the inevitable momentary dip with x3d.

That said, the 7800x3d has been an excellent avenue for gamers who don't care about ST/MT scores plus they might not even notice the dips anyway. They just want to drop in a CPU and play their games and instantly get ~90% of the top possible performance without any tuning on their end. I personally recommend X3D AM5 systems to friends too, for its ease of use and (recently) lower platform cost.

AMD's trump card has been their cache technology, but their normal chips have been falling behind consistently, gen to gen. People often criticize that the intel chips draw an insane amount of power. Yes, this is true. For most of us though, we are playing games, and in most games intel chips do not draw significantly high power. Contrary to that, while pulling more power, more often than not it can be running cooler than its AMD counterparts due to better IHS design, bigger and centered die. In games, we are talking perhaps at most a 50w deviation from the various classes of chips between Intel and AMD.

If you're arguing about running costs:
50w x 4 hours of gaming daily x 31 days = 6200w / 6.2kW
6.2kW x $0.3 = $1.86/m / $22.32

So perhaps running an AMD saves you ~$23 more per year. Good for you.

For those that want both gaming and productivity performance, it's either the 7950x3d or 13900k/14900k. The intel chips are cheaper, perform equally, and both chips run hot. Intel because of its power use at full load, AMD because of its algorithm seeking max clock performance at max temps. Dual CCD chips have their own issues as well, such as deciding which CCD to use etc.

I'm no shill for any company, I only gun for top spot, whoever it is. And for multiple generations now, for the enthusiasts with knowledge, it has always been Intel. If one day the tuning top dog is AMD, you'll see me on the leader boards with an AMD platform.

I understand preferring one company over the other, and that's totally fine. But to blanket cover an entire company just for your superiority complex is subpar behavior. Do better.

 

[Mach3.2]

keeping idle power consumption low, i.e. 7-10w idle on intel

Actually intel has been doing this with their monolithic designs since forever. Even AMD monolithic designs can do this, only their chiplet designs idle at 30-40w. It's worse on AM4 if you overclock your FCLK to above 1600MHz.

imo E cores are a crutch Intel needed to compete with AMD. Even Intel indirectly admit it's a crutch if you look at their server and workstation products. If it's really that good, why don't they have E cores on sapphire rapids?


One can only hope their main architecture can compete with AMD enough that they don't have to do the E core nonsense in a few generations time.

 

[matique]

Actually intel has been doing this with their monolithic designs since forever. Even AMD monolithic designs can do this, only their chiplet designs idle at 30-40w. It's worse on AM4 if you overclock your FCLK to above 1600MHz.

imo E cores are a crutch Intel needed to compete with AMD. Even Intel indirectly admit it's a crutch if you look at their server and workstation products. If it's really that good, why don't they have E cores on sapphire rapids?


One can only hope their main architecture can compete with AMD enough that they don't have to do the E core nonsense in a few generations time.

oh yeah definitely, i would have loved a 10-12 pcore setup, but it is what it is. Crutch or no crutch, it does help with MT perf if you do need it. In an ideal world, it would have been nice to have intel pcore perf, scale it higher core count, add AMD's vcache, stick with intel's mono design w/ better IHS, and improved memory controller.

But one can only dream and hope competition breeds better chips like this. Both companies need each other to push one another, so I'm happy as long as they're closely competing.

 

[Mach3.2]

oh yeah definitely, i would have loved a 10-12 pcore setup, but it is what it is. Crutch or no crutch, it does help with MT perf if you do need it. In an ideal world, it would have been nice to have intel pcore perf, scale it higher core count, add AMD's vcache, stick with intel's mono design w/ better IHS, and improved memory controller.

But one can only dream and hope competition breeds better chips like this. Both companies need each other to push one another, so I'm happy as long as they're closely competing.

I mean what I was wishing for is essentially Sapphire Rapids-WS but 12c or 16c max in a consumer platform.

But am I willing to pay the premium for SR-WS? Honestly no. So i guess i can continue crying in the mean time.


Anyway Intel have some LGA1700 Xeon parts listed on their site so that might be a source for some LGA1700 8 P core only chips when it launches in 2024.

 

[matique]

I mean what I was wishing for is essentially Sapphire Rapids-WS but 12c or 16c max in a consumer platform.

But am I willing to pay the premium for SR-WS? Honestly no. So i guess i can continue crying in the mean time.


Anyway Intel have some LGA1700 Xeon parts listed on their site so that might be a source for some LGA1700 8 P core only chips when it launches in 2024.

Yeah but the thing with spr is that it won't clock as high so far I'm content with my setup. 6ghz pcore, 4.6ghz ecore, HT off, 1.35v. 8000c34 ram. Stable as can be, runs super cool, it just works

 

[Encrypted11]

Intel might have a better design with probably 10-12 P-Core + 1-2 E Core cluster. X3D VCache are simply just an extra SRAM tile on the CPU giving it more fast on-chip memory access. Similar to what Broadwell-E has done but it adds cost to the chip, which is not a proposition Intel’s TMG or its board can accept. Intel’s board after Paul O.’s time especially closer to Brian K.’s time has a lot less engineers and a lot more MBAs instead.

I personally think the increase in E-Core count is a by product of raving for high CineBench MT scores which is not a meaningful real world performance yardstick in many cases.

Then again, I still think Intel underperforming expectations considering the huge trove of Intellectual Property it is sitting on are just a by product of an under performing management.

They were and *are* still the technical excellence company (Think IBM’s contribution to the PC long time ago) that is driving standards like PCI-SIG, JEDEC, WiFi/BT Alliance (to a lesser degree) and many more forward. These stuff did materially give us all a better computer.

Give credit when it is due. Corporations are not sports teams!

 

[ragnarok95]

Son: Mummy i want fast fast computer.
Mummy: Okie.
Mummy go sim lim.
Mummy: hi, i want to build a fast fast PC.
Sales: intel 14900K, best from intel.
Mummy leave the shop with a smile.

 

[matique]

Intel might have a better design with probably 10-12 P-Core + 1-2 E Core cluster. X3D VCache are simply just an extra SRAM tile on the CPU giving it more fast on-chip memory access. Similar to what Broadwell-E has done but it adds cost to the chip, which is not a proposition Intel’s TMG or its board can accept. Intel’s board after Paul O.’s time especially closer to Brian K.’s time has a lot less engineers and a lot more MBAs instead.

I personally think the increase in E-Core count is a by product of raving for high CineBench MT scores which is not a meaningful real world performance yardstick in many cases.

Then again, I still think Intel underperforming expectations considering the huge trove of Intellectual Property it is sitting on are just a by product of an under performing management.

They were and *are* still the technical excellence company (Think IBM’s contribution to the PC long time ago) that is driving standards like PCI-SIG, JEDEC, WiFi/BT Alliance (to a lesser degree) and many more forward. These stuff did materially give us all a better computer.

Give credit when it is due. Corporations are not sports teams!

Very much agreed. Imo Intel has catching up to do with AMD in terms of catering to the audience. Like you said, get a 10 pcore chip in maybe with 2-4 ecores to handle simple background tasks, so a 10p4e24t cpu.

But the future seems exciting with both AMD and Intel, both going into the next round with much revised chips.

Considering next round of GPUs also potentially making a big leap, it's important not to be cpu bound.

 

[ragnarok95]

IIRC, SD also reduce "E" core in this yr 8Gen3.

 

[watzup_ken]

Actually intel has been doing this with their monolithic designs since forever. Even AMD monolithic designs can do this, only their chiplet designs idle at 30-40w. It's worse on AM4 if you overclock your FCLK to above 1600MHz.

imo E cores are a crutch Intel needed to compete with AMD. Even Intel indirectly admit it's a crutch if you look at their server and workstation products. If it's really that good, why don't they have E cores on sapphire rapids?


One can only hope their main architecture can compete with AMD enough that they don't have to do the E core nonsense in a few generations time.

I feel the "efficient" cores are basically to make up for the lack of threads, and also to reduce cost + die space. Clearly one won't need 16 E-cores to efficiency. Anything more than 8 E-cores is excessive. A formidable challenge that Intel is going to face when going down this P/E core path, is the software to manage the cores. Windows is a mess of an OS that have little focus on hardware optimization simply because there are too much hardware combinations that also makes it impossible for them to do a proper job of optimizing the software with the hardware.

 

[Phen8210]

I feel the "efficient" cores are basically to make up for the lack of threads, and also to reduce cost + die space. Clearly one won't need 16 E-cores to efficiency. Anything more than 8 E-cores is excessive. A formidable challenge that Intel is going to face when going down this P/E core path, is the software to manage the cores. Windows is a mess of an OS that have little focus on hardware optimization simply because there are too much hardware combinations that also makes it impossible for them to do a proper job of optimizing the software with the hardware.

I think it's everything mentioned by me, others, and including what you said. Basically, sooner or later people will learn that these e-cores are nothing but glorified low-performance cores. The ones more familiar and caught up already knew of course.

I'm sorry to say that the entire point of its existence is to pull off a marketing stunt that has caused nothing but trouble for both users and related vendors, but its true.

 

[watzup_ken]

I think it's everything mentioned by me, others, and including what you said. Basically, sooner or later people will learn that these e-cores are nothing but glorified low-performance cores. The ones more familiar and caught up already knew of course.

I'm sorry to say that the entire point of its existence is to pull off a marketing stunt that has caused nothing but trouble for both users and related vendors, but its true.

I feel there are merits to having e-cores just as we have seen with mobile SOCs. However, Intel is charging quite a lot for what is mostly E-cores in their higher end chips, with essentially the same number of more sophisticated and expensive cores. With Meteor Lake, this is going to regress to just 6 P-cores, and again, more E-cores spamming.

 

[matique]

I think it's everything mentioned by me, others, and including what you said. Basically, sooner or later people will learn that these e-cores are nothing but glorified low-performance cores. The ones more familiar and caught up already knew of course.

I'm sorry to say that the entire point of its existence is to pull off a marketing stunt that has caused nothing but trouble for both users and related vendors, but its true.

There are some merits to ecores, but generally I would agree with your statement that it is glorified low performance cores. That said, it's not like there's anything we can do nor could we argue much about it since it clearly works, since games now generally perform better with ecores turned on vs off, unlike say during launch period of 12th gen where it's better to have ecores disabled.

Like I mentioned before, even AMD is looking into performance, efficiency and acceleration cores in the future. A small indication of this is already in production with their 7040 Phoenix 2 processor, where there 4x zen4 cores and 2x zen4c cores, with zen4c cores being more efficient at lower power due to density and thus being clocked lower. Curiously AMD does not have a dedicated thread director like intel's to allocate which tasks head to which core.

Should AMD move forward with the concept, would you be as critical of the hybrid architecture? Games have to play catch up, so do programs, but once they get it right it should be alright for the end consumer. I would always prefer more pcores but due to the node design and various other issues, Intel is hesitant to do so.

At the end of the day, we choose the platform that suits our needs the most. For me personally, the i9 chips provide the best ST perf, best OC scaling, and if needed, best MT perf even though it's cheesing it with ecore spamming.

 

[Mach3.2]

There are some merits to ecores, but generally I would agree with your statement that it is glorified low performance cores. That said, it's not like there's anything we can do nor could we argue much about it since it clearly works, since games now generally perform better with ecores turned on vs off, unlike say during launch period of 12th gen where it's better to have ecores disabled.

Like I mentioned before, even AMD is looking into performance, efficiency and acceleration cores in the future. A small indication of this is already in production with their 7040 Phoenix 2 processor, where there 4x zen4 cores and 2x zen4c cores, with zen4c cores being more efficient at lower power due to density and thus being clocked lower. Curiously AMD does not have a dedicated thread director like intel's to allocate which tasks head to which core.

Should AMD move forward with the concept, would you be as critical of the hybrid architecture? Games have to play catch up, so do programs, but once they get it right it should be alright for the end consumer. I would always prefer more pcores but due to the node design and various other issues, Intel is hesitant to do so.

At the end of the day, we choose the platform that suits our needs the most. For me personally, the i9 chips provide the best ST perf, best OC scaling, and if needed, best MT perf even though it's cheesing it with ecore spamming.

I agree on the points being made.

In order for E cores to be acceptable for me, both P and E cores must support the same instruction sets and the scheduler must be intelligent enough to handle the disparity in core performance. And of course the software have to support that too.

Actually if both E core and P cores have uniform instruction sets, I believe the scheduler woudn't have much difficulties since AMD Zen 2/3 CPUs are essentially dealing with a very similar issue with their dual CCD chips; the 2nd CCD isn't as strong as the first CCD, albeit the difference isn't very large.

As you also mentioned, I do see improvements in the scheduler being reported as compared to the early days of Intel 12th gen, so once again, credit where it's due.

 

[matique]

my avx512

 

[watzup_ken]

There are some merits to ecores, but generally I would agree with your statement that it is glorified low performance cores. That said, it's not like there's anything we can do nor could we argue much about it since it clearly works, since games now generally perform better with ecores turned on vs off, unlike say during launch period of 12th gen where it's better to have ecores disabled.

Like I mentioned before, even AMD is looking into performance, efficiency and acceleration cores in the future. A small indication of this is already in production with their 7040 Phoenix 2 processor, where there 4x zen4 cores and 2x zen4c cores, with zen4c cores being more efficient at lower power due to density and thus being clocked lower. Curiously AMD does not have a dedicated thread director like intel's to allocate which tasks head to which core.

Should AMD move forward with the concept, would you be as critical of the hybrid architecture? Games have to play catch up, so do programs, but once they get it right it should be alright for the end consumer. I would always prefer more pcores but due to the node design and various other issues, Intel is hesitant to do so.

At the end of the day, we choose the platform that suits our needs the most. For me personally, the i9 chips provide the best ST perf, best OC scaling, and if needed, best MT perf even though it's cheesing it with ecore spamming.

I think if you read up on the Zen 4C, architecturally, it is the same as the normal Zen 4 cores with some cutbacks to make it less power hungry. So you are essentially still getting a fairly high performing E-core. Intel's current E-core may be around Sky/ Coffee Lake performance, but it is nowhere near the P-core performance. Hence the need to spam them in great quantities. Again, I am mostly unhappy because they are selling these supposed cheap cores at a premium just because you get 16 of them in an i7/i9 processor. On the mobile side of things, it looks even worst because you can get as low as 6 P-cores on an i7 mobile processor, slower iGPU and yet cost more than what AMD is offering.

 

[Phen8210]

Intel designed their chip knowing how Computers and OS function, and that such a design would be unfavorable to both users and vendors.

Devs for OS and applications have more important tasks at hand than wasting their resource on something that isn't even a technological advancement or a straightforward solution.

This is in contrast to Nvidia's introduction of features like Ray Tracing and AI Upscaling with DLSS to their hardware, which is an actual technology advancement and yet remains relatively easy to implement into applications.

 

[Mach3.2]

my avx512

Ya I don't even get it with my late production P cores only 12500

Thanks Steve.

 

[matique]

I think if you read up on the Zen 4C, architecturally, it is the same as the normal Zen 4 cores with some cutbacks to make it less power hungry. So you are essentially still getting a fairly high performing E-core. Intel's current E-core may be around Sky/ Coffee Lake performance, but it is nowhere near the P-core performance. Hence the need to spam them in great quantities. Again, I am mostly unhappy because they are selling these supposed cheap cores at a premium just because you get 16 of them in an i7/i9 processor. On the mobile side of things, it looks even worst because you can get as low as 6 P-cores on an i7 mobile processor, slower iGPU and yet cost more than what AMD is offering.

yup yup architecturally the same, it's just smaller = denser = hotter = not clocked as high. IPC remains the same. I merely referenced it because it shows AMD is looking at segmenting cores as well, not just Intel and SD. They will probably move forward with more segmentation in the future. We may not like it, but it is what it is.

And yeah for laptop I just use AMD. Best use case for me. Currently using a Lenovo Slim 7 Pro with a 7840HS. Plenty of cores and fast enough for laptop use. Great value too got it at $1200. 16gb D5 6400mhz, 512gb stock gen4 ssd but i added an old 1tb gen3 drive. 90hz 2560x1600 screen.

Ya I don't even get it with my late production P cores only 12500

Thanks Steve.

heh i had it on my 12900k, was fun to do Ycruncher benches on it!

 

[RockaRolla]