• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Intel 12th / 13th gen CPU latency

BFG10K

BFG10K

Lifer
This is a brand new video. He's saying 12th / 13th gen Intel CPUs had their northbridge moved back off the CPU which increases latency visibly compared to older CPUs.


Has this been discussed or observed elsewhere?
 
northbridge out of the CPU? what? like in it's on cluster with higher latency interconnect still within the same die or something?
I have no idea, it sounds super weird.
 
First mistake, larger nm does not equal slower, it's larger and maybe needs more power but the ability to switch fast between high and low has nothing to do with how big each transistor is.

Second mistake, for the amount of lag that he is showing the northbridge would have to be on a different computer, it's completely ridiculous to think that it could add that much lag.

Third mistake, calling the northbridge a driver, drivers are software while the hardware is called a controller, if you want to call it anything other than its real name.


Also we have very high res and detailed alder lake die shots that show the memory and pci controllers still being inside the CPU.
Yes, the CPU cores could be an individual tile but I don't think that the northbridge ever was a part of the actual cores, I mean even he says that much. The whole area on the left of the cores is all IO.
Die_Shot_Blank_com.jpg
 
So I don't think he has the technical knowledge to really make a video like this as he is just wrong on the whole northbridge thing and the source of this, "latency". As far as his getting confirmation from people at Computex, either he is talking to people who don't know what they are talking about or is misunderstanding what they are saying.

With that said, the behavior should be investigated if multiple people are reporting the same experience. As for my own anecdotal experience, I have actually experienced what he is talking about on my work laptop with Alderlake. Since it's a work laptop I can't really do anything to investigate it but it will randomly kind of feel at times like I am running off an old spinner drive even though there is only an SSD in the laptop.
 
Diagnosis is wrong. But I can't even tell what he is trying to blame it on.

Why not try basic problem solving before tilting at windmills?
 
Very bad way to approach this: I experience something on system X, experience something else on system Y, therefore the problem is this exact issue Z. No objective measurements, no attempt to isolate the cause. The least effort necessary here is to measure the difference. If it's a hardware problem, it will show up in synthetic tests that bombard the I/O system with single threaded random requests. Then we can start asking why.
 
coercitiv said:
Very bad way to approach this: I experience something on system X, experience something else on system Y, therefore the problem is this exact issue Z. No objective measurements, no attempt to isolate the cause. The least effort necessary here is to measure the difference. If it's a hardware problem, it will show up in synthetic tests that bombard the I/O system with single threaded random requests. Then we can start asking why.
Click to expand...
Yet another case of a tech influencer getting the details wrong. What's up with today's tech "journo" climate? Everyone with a modicum of influence just seems to be no more than a hardware enthusiast masquerading as someone who understands this stuff competently. There's just so much misinformation and/or low-effort investigative work. For every Gamer's Nexus, there's likely a dozen one of them (I lump the bog standard rumor mill websites and Youtubers in there as well).
 
I had online storage gdrive mounted as drives in windows and sure as soon they throttle me for "abuse" I would wait for minutes in my explorer, same here its OS and slow drive issue, and here both AMD and Intel has dedicated one or more slots from nvme
He also know AMD first kicked northbridge out of cpu units, and he do zero test, and only -profanity removed- he slow is his broken windows 😀

Profanity is not allowed in the technical forums.

Daveybrat
AT Moderator
 
Last edited by a moderator:
PCI-E lanes are still on the CPU and not on the board in most cases.
That means the Northbridge is STILL on the CPU.

In old boards and systems, the Northbridge handled everything but the actual CPU compute.
They slowly started moving stuff off the North and South bridge, and we lost the northbridge entirely.

It first started with the Memory controller being moved onto the cpu, then the PCI-E lanes.
As long as these two are still in the CPU, the Northbridge is on the CPU.

There are "Fake" northbridge that some enterprise class boards uses, usually in multi cpu applications, like when you have a 4S board. But this is way beyond what consumers will ever get, as i have yet to see a consumer need a 4 cpu board.

Hell, even EYPC doesn't have a 4 CPU board as i hear the interconnects for 4 cpu's is just a technical nightmare on EYPC.
 
Last edited:
Something's definitely going on here with the amount of people that reported similar problems to him.

I don't think it's the Northbridge but it might be a BIOS issue. Or it could be E-core latency /scheduling; would be interesting to see him re-test with them disabled.
 
One thing that could impact responsiveness is context switching. These happen anytime a thread moves between cores or some application makes a system call through Win32 API which causes a switch from user mode to kernel mode and back again. But even there, 13900K is faster than 10900K as shown here: https://openbenchmarking.org/test/pts/ctx-clock&eval=e73e6187dbf9f015fb568fe82670e0c6cdb6cacf

Another way to benchmark this overhead is to run some database workload on the main system and then inside a VM on the same system. Calculate the performance hit in percentage and that's the context switching overhead. CPU with the lower percentage hit wins.
 
Context switching has improved and thus we're at a loss here. chiming in with bfg here it may be down to microsoft a/b testing performance factors behind the scenes on wind 10 and 11 systems without user knowledge. it's not that crazy to consider. in my own testing with alder lake and rapture lake systems I find them much snappy compared to skylake era processors. i do not believe there will be a conclusive finale to this puzzling question and blame will go around.
 
I only have 11th gen Intel parts (Tiger Lake and Rocket Lake) so can't test this out.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top