SNAP - The Intel CPU Latency Story (Part 1 of 2)

The Latency Enthusiast's Conundrum: A Deep Dive into Intel CPU Architecture

As a latency enthusiast, I have always been fascinated by the intricacies of CPU architecture and its impact on system performance. Over the years, I have experimented with various CPU configurations, including switching from higher-core-count CPUs to lower-core-count processors with older silicon. This approach has allowed me to achieve better performance in certain tasks, which has led me to question the current trends in CPU design.

One trend that stands out is the need for closer proximity between the CPU and IO drivers. Historically, advancements in silicon technology have enabled performance enhancements by moving these components closer together on the same piece of silicon as the CPU. However, with the introduction of 12th and 13th gen Intel CPUs, this approach has been reversed. For the first time in Intel's history, the system agent that was previously located on the same piece of silicon as the CPU has now been moved off the CPU itself. This change is directly responsible for issues experienced by users running Windows 10 and 11 with the i9-12900K processor, including noticeable latency and slowdowns.

I initially attributed this issue to the ecores, which were introduced in Intel's hybridized ring bus architecture that merged the classical Intel ring bus with the mesh architecture. To test my theory, I disabled ecores and set static PCIe speeds and voltages, but the lag persisted. Further investigation revealed that snapping into Windows, even at fast refresh rates, would still cause noticeable delays. Witnessed firsthand by Tech yes City team members, including GPU Tech tuber Marco, this issue is not an isolated phenomenon. Even simple tasks like searching for an MP3 file or an MP4 movie file, or dragging and dropping files between programs, were affected.

The problem was so pronounced that I began to doubt my own perception of latency, wondering if it was indeed a figment of my imagination. However, after speaking with multiple individuals at Computex, including representatives from Intel themselves, the truth became clear: this is not just a case of being overcautious; it's a genuine problem caused by the IO driver being moved directly off the CPU.

This development brings back memories of past discussions about CPU architectures and their performance characteristics. In particular, I recall debates surrounding the Intel Core i9-10900K vs. its Ryzen 5 3950X counterpart from AMD. At the time, opinions were divided on whether the Intel architecture felt snappier due to its full ring bus architecture and fully integrated IO driver on the die, versus the separate die for the platform chipset hub, which used a global foundry's 12 nanometer technology. While the debate was contentious, it was never truly resolved.

Now that we have identified the root cause of the issue – the separation of the IO driver from the CPU core – I believe this trend will continue and worsen in the future. In fact, I expect AMD to follow suit with its next-generation Ryzen CPUs, potentially further exacerbating the problem.

For those still using the i9-10900K and considering upgrading, I would advise caution. While Intel's newer architectures may offer improved performance in some areas, the latency issues highlighted here may persist or worsen as the technology continues to evolve. In fact, I plan to upgrade to 10th gen CPUs soon and will be creating content around my experience. Stay tuned for part two of this series, where we'll delve into the future of CPU architecture and its implications for latency enthusiasts like myself.

In conclusion, the story of Intel's CPU architecture is one of trial and error, with each new generation bringing about both improvements and setbacks. While it may seem counterintuitive to move the IO driver away from the CPU core, this trend has brought attention to a critical aspect of system performance: latency. As we continue down this path, it will be essential to monitor developments closely and share our findings with the community.

The Future of Intel and AMD CPUs

As we conclude part one of our exploration into the world of latency and CPU architecture, I want to emphasize that the story is far from over. In fact, I firmly believe that things are about to get worse before they get better. The separation of the IO driver from the CPU core has already had a significant impact on system performance, and I predict that this trend will continue with AMD's next-generation Ryzen CPUs.

While Intel has made strides in recent years, its decision to move the system agent off the CPU core has raised concerns among enthusiasts like myself. As we head into the future, it's essential to understand what this means for users and how we can adapt to these changes.

One thing is clear: AMD will likely follow suit with its own innovations, potentially exacerbating the problem of latency in CPUs. The question remains whether Intel will be able to regain control over system performance through software or architectural advancements.

The Debate Continues

In recent years, debates surrounding CPU architectures have been contentious, but never truly resolved. As we continue down this path, it's essential to keep an open mind and consider multiple perspectives. While the i9-10900K vs. Ryzen 5 3950X debate may seem like a relic of the past, its implications for future generations are still relevant.

The Future of Latency Enthusiasts

As we move forward, I believe it's crucial to stay vigilant and monitor developments closely. For latency enthusiasts like myself, this means being prepared to adapt to changing circumstances and share our findings with the community. By working together, we can ensure that our systems continue to perform at their best, even as technology evolves.

The story of Intel's CPU architecture is one of trial and error, with each new generation bringing about both improvements and setbacks. As we navigate this complex landscape, it's essential to remain informed, share our knowledge, and support one another in our quest for optimal system performance.

Stay tuned for part two of our series, where we'll delve deeper into the world of CPU architecture and explore its implications for latency enthusiasts like myself.

"WEBVTTKind: captionsLanguage: enas I was walking around Taiwan at the start of June this year for computex I was trying to find out as much information as possible about gpus and their manufacturing costs especially in recent light of the release of the RTX 4060 TI however what I actually got was a lot of complaints about CPUs and motherboards from the people on the showroom floor however more about this after today's video sponsor never pay full price for Windows 10 or 11 again with today's video sponsor SCD Keys you can get activated for as little as $15 using that coupon bftc links in the description below so interesting was this information divulged that I decided to make a two-part series on this topic though we will break down GPU cost in future videos which I do have very strict private information in relation to this but that is something where I want to cast out a fishing net so to speak and see the industry among fellow YouTubers who are telling total BS or who is giving you accurate information that story however will take some time to brew then let's talk about CPUs and in particular Intel where this story goes all the way back to the first generation of i7 CPUs where on the motherboard you had what was known as a North Bridge and a South Bridge two separate pieces of silicon fused onto the motherboard then you had at the center the CPU which was the third piece of silicon sold by Intel and if you were to build a computer at this time it would be comprised of say something like an i7 920 and an x58 motherboard the CPU however was the most expensive piece of silicon and had the latest groundbreaking technological advancements second was the North Bridge which was made of Cheaper silicon but not as cheap as the South Bridge generally which in this priority the North Bridge was responsible for handling pcie the input output driver and we're going to stop quickly there it's built into the name x58 ioh ioh stands for input output Hub this controller has to still perform quite fast which is why on x58 in particular when you raise the speed of this Hub both itself and through the interconnect speeds you can notice a visible speed increase or to put it in more simple terms as I would say the system felt more snappier however even at its Max speeds it was starting to hold Intel back and with the introduction of Sandy Bridge that's Intel second generation Intel completely removed this North Bridge and integrated it into the CPU itself now calling it the system agent or the uncore a not CPU CES but still attached to the CPU and this is the key right here it allowed the CPU to not only run faster but also reduce the latency a king talking topic of today's video where now the iio driver or ioh was located directly on the same piece of silicon as the CPU yes it was the most expensive piece of silicon but this was a game-changing move that saw Sandy Bridge AKA and i72600 K for example be one of the most if not the most iconic CPU releases of all time the South Bridge on the other hand responsible for handling USB Nick audio and hard drive throughput was the slowest piece of silicon offered from Intel and came to be renamed as we know it as the PCH or platform chipset Hub and still to this day is using slower silicon than the CPU cores actually nowadays it's usually a generation of silicon behind with an I9 13900 K using 10 nanm silicon AKA Intel 7 and a z790 chipset using 14 nanm silicon on the PCH or the chip that is integrated into the motherboard so with that backstory complete it's actually very important because it leads us now to the problem and a forewarning this may not be a problem for the majority of people I am just quite the latency Enthusiast here at Tech yes City and delving into the topic on numerous occasions in the past has led me to even ditch higher core newer CPUs in favor of older lower core count zons for example and getting a better experience at least for what I do in the process though back to the topic at hand the backstory here displays a trend that took place in the CPU industry that is in order for performance to become greater latency being such metric silicon had to move closer to the CPU and use better silicon all around not just on the CPU itself and here is where with the introduction of 12th and now 13th gen Intel CPUs something has reversed especially for the first time in Intel's history of making CPUs that IO driver functionality of the system agent that was located on the same piece of silicon as the CPU has now and I was told by multiple Anonymous sources at copex has now been moved off the CPU and this is directly responsible for issues on Windows 10 and 11 when using an i 92900 k for example and Performing even simple tasks as for instance searching for an MP3 file or an mp4 movie file or even dragging and dropping files between different programs I initially I thought this latency issue or this slowdown of performance was due to the ecores and if you guys didn't know on 12th and 13th gen Intel has hybridized the classical Intel ring bus AK the peores and merged them with ecores which came from the mesh architecture AKA they all merge together in a complex architecture however I tried on top of this disabling ecores and setting static peor speeds and voltages all to no avail the lag was still there and when snapping in Windows fast enough we had Witnesses here at Tech City that could see the problem firsthand for instance Marco the GPU Tech tuber who is almost in the comment section of every single Tech yes video here he got to witness a tech yes video being edited in real time and he would describe it as snap yes City I really do love latency or put it this way low latency a recent example of this too would be making my kids some practice Pokémon cards and even Microsoft Word for instance just bugged out there is something seriously wrong with 12th and 13th gen from Intel and it's not the thread director which was heavily updated in collaboration with Intel themselves and I thought before going to Taiwan that maybe this was all in my head perhaps I was crazy I mean my wife certainly likes to think I'm crazy so this could just be another one of those things the 2 to 3 second delays on searching the bugs incurred when dragging and dropping files it was all just a figure of my imagination until finally on that showroom floor at computex I got answers from multiple people saying no it wasn't in your head and it's actually a problem and it has to do once again with the io driver being moved directly off the CPU which is why circling back to a few years ago if you dig deep enough on the internet you'll hear opinions mine included talking about how Intel CPUs felt snappier than their rise and counter parts for instance take the I9 10900 K full ring bus architecture fully integrated IO driver on the die and the same cicon being used on the platform chipset Hub as the CPU versus their ryzen counterparts say a ryzen 5 3950x at the time which used actually better silicon from tsmc on the CPU cores themselves but used a global foundry's 12 nanometer for the io driver IMC and pcie controllers which was detached from the CPU cores themselves on a separated die connected via what is known as the infinity fabric that snappiness debate which in the past has been met with controversy now finally for me at least had an answer and I can put the debate to rest so if you're still on an I9 10900k and you're cautious about upgrading I personally wouldn't in fact I'm going to be downgrading to 10th gen very soon and I'll be making some content around it though with that aside guys part one has come to a close here today you may be wondering there was a lot that was answered here do we need a part two the answer is definitely yes because we've only identified the problem here in part one in part two we're going to talk about the future of Intel and AMD CPUs and it's only unfortunately in my opinion going to get worse so stay tuned if you guys like this content especially if you're a latency driven hothead also do let us know in the comments section below your experiences with both Intel and AMD CPU architectures love reading your thoughts and opinions as always and I'll see you in the next installment very soon peace out for now byeas I was walking around Taiwan at the start of June this year for computex I was trying to find out as much information as possible about gpus and their manufacturing costs especially in recent light of the release of the RTX 4060 TI however what I actually got was a lot of complaints about CPUs and motherboards from the people on the showroom floor however more about this after today's video sponsor never pay full price for Windows 10 or 11 again with today's video sponsor SCD Keys you can get activated for as little as $15 using that coupon bftc links in the description below so interesting was this information divulged that I decided to make a two-part series on this topic though we will break down GPU cost in future videos which I do have very strict private information in relation to this but that is something where I want to cast out a fishing net so to speak and see the industry among fellow YouTubers who are telling total BS or who is giving you accurate information that story however will take some time to brew then let's talk about CPUs and in particular Intel where this story goes all the way back to the first generation of i7 CPUs where on the motherboard you had what was known as a North Bridge and a South Bridge two separate pieces of silicon fused onto the motherboard then you had at the center the CPU which was the third piece of silicon sold by Intel and if you were to build a computer at this time it would be comprised of say something like an i7 920 and an x58 motherboard the CPU however was the most expensive piece of silicon and had the latest groundbreaking technological advancements second was the North Bridge which was made of Cheaper silicon but not as cheap as the South Bridge generally which in this priority the North Bridge was responsible for handling pcie the input output driver and we're going to stop quickly there it's built into the name x58 ioh ioh stands for input output Hub this controller has to still perform quite fast which is why on x58 in particular when you raise the speed of this Hub both itself and through the interconnect speeds you can notice a visible speed increase or to put it in more simple terms as I would say the system felt more snappier however even at its Max speeds it was starting to hold Intel back and with the introduction of Sandy Bridge that's Intel second generation Intel completely removed this North Bridge and integrated it into the CPU itself now calling it the system agent or the uncore a not CPU CES but still attached to the CPU and this is the key right here it allowed the CPU to not only run faster but also reduce the latency a king talking topic of today's video where now the iio driver or ioh was located directly on the same piece of silicon as the CPU yes it was the most expensive piece of silicon but this was a game-changing move that saw Sandy Bridge AKA and i72600 K for example be one of the most if not the most iconic CPU releases of all time the South Bridge on the other hand responsible for handling USB Nick audio and hard drive throughput was the slowest piece of silicon offered from Intel and came to be renamed as we know it as the PCH or platform chipset Hub and still to this day is using slower silicon than the CPU cores actually nowadays it's usually a generation of silicon behind with an I9 13900 K using 10 nanm silicon AKA Intel 7 and a z790 chipset using 14 nanm silicon on the PCH or the chip that is integrated into the motherboard so with that backstory complete it's actually very important because it leads us now to the problem and a forewarning this may not be a problem for the majority of people I am just quite the latency Enthusiast here at Tech yes City and delving into the topic on numerous occasions in the past has led me to even ditch higher core newer CPUs in favor of older lower core count zons for example and getting a better experience at least for what I do in the process though back to the topic at hand the backstory here displays a trend that took place in the CPU industry that is in order for performance to become greater latency being such metric silicon had to move closer to the CPU and use better silicon all around not just on the CPU itself and here is where with the introduction of 12th and now 13th gen Intel CPUs something has reversed especially for the first time in Intel's history of making CPUs that IO driver functionality of the system agent that was located on the same piece of silicon as the CPU has now and I was told by multiple Anonymous sources at copex has now been moved off the CPU and this is directly responsible for issues on Windows 10 and 11 when using an i 92900 k for example and Performing even simple tasks as for instance searching for an MP3 file or an mp4 movie file or even dragging and dropping files between different programs I initially I thought this latency issue or this slowdown of performance was due to the ecores and if you guys didn't know on 12th and 13th gen Intel has hybridized the classical Intel ring bus AK the peores and merged them with ecores which came from the mesh architecture AKA they all merge together in a complex architecture however I tried on top of this disabling ecores and setting static peor speeds and voltages all to no avail the lag was still there and when snapping in Windows fast enough we had Witnesses here at Tech City that could see the problem firsthand for instance Marco the GPU Tech tuber who is almost in the comment section of every single Tech yes video here he got to witness a tech yes video being edited in real time and he would describe it as snap yes City I really do love latency or put it this way low latency a recent example of this too would be making my kids some practice Pokémon cards and even Microsoft Word for instance just bugged out there is something seriously wrong with 12th and 13th gen from Intel and it's not the thread director which was heavily updated in collaboration with Intel themselves and I thought before going to Taiwan that maybe this was all in my head perhaps I was crazy I mean my wife certainly likes to think I'm crazy so this could just be another one of those things the 2 to 3 second delays on searching the bugs incurred when dragging and dropping files it was all just a figure of my imagination until finally on that showroom floor at computex I got answers from multiple people saying no it wasn't in your head and it's actually a problem and it has to do once again with the io driver being moved directly off the CPU which is why circling back to a few years ago if you dig deep enough on the internet you'll hear opinions mine included talking about how Intel CPUs felt snappier than their rise and counter parts for instance take the I9 10900 K full ring bus architecture fully integrated IO driver on the die and the same cicon being used on the platform chipset Hub as the CPU versus their ryzen counterparts say a ryzen 5 3950x at the time which used actually better silicon from tsmc on the CPU cores themselves but used a global foundry's 12 nanometer for the io driver IMC and pcie controllers which was detached from the CPU cores themselves on a separated die connected via what is known as the infinity fabric that snappiness debate which in the past has been met with controversy now finally for me at least had an answer and I can put the debate to rest so if you're still on an I9 10900k and you're cautious about upgrading I personally wouldn't in fact I'm going to be downgrading to 10th gen very soon and I'll be making some content around it though with that aside guys part one has come to a close here today you may be wondering there was a lot that was answered here do we need a part two the answer is definitely yes because we've only identified the problem here in part one in part two we're going to talk about the future of Intel and AMD CPUs and it's only unfortunately in my opinion going to get worse so stay tuned if you guys like this content especially if you're a latency driven hothead also do let us know in the comments section below your experiences with both Intel and AMD CPU architectures love reading your thoughts and opinions as always and I'll see you in the next installment very soon peace out for now bye\n"