Why can't Intel do this?
>When AMD couldn't profitably put more cache on the same 2D surface as the CPU cores, they went vertical.

>The "3D" V-Cache in a chip like the 7800X3D (and its future successors) is a revolutionary piece of packaging technology.

>What it is: AMD literally stacks an extra layer of L3 cache directly on top of the processor core.

>Why it's a game-changer: This allows them to have a gigantic 96MB+ L3 cache. For tasks that are sensitive to memory latency—like gaming—this is a MONUMENTAL advantage. The CPU almost never has to wait for data from RAM because it's nearly always available in this huge, stacked cache. This provides a massive performance boost in games, even if the clock speed is slightly lower than non-3D chips.

>>105730963 (OP)
AMD has to use a gimmick to not fall behind because they can't make a good memory controller

>>105730980
All I know is that the 9800X3D gives 1500 fps in valorant and the best intel cpu gives like 600 fps

>>105730963 (OP)
>literally stacks an extra layer of L3 cache directly on top of the processor core
Wow, that makes it sound so easy and not something that requires hundreds of millions if not billions of dollars and multiple years of R&D and a massive shift in design just to play catch up with AMD.

>>105730984
The gap closes with high frequency lower latency ram
One thing intel does do better is memory controller stability
Zen 5 starts shitting the bed at 6500hz

>>105730963 (OP)
intel embraced ddr5 and they dont have to waste money on adding bajilobytes of cache lmao

>>105731012
One of my PCs basically has a world record overclock speed on Samsung B-Die ram. Not sure how much it matters but its at least noticeable

>>105730963 (OP)
>The CPU almost never has to wait for data from RAM because it's nearly always available in this huge, stacked cache
ha-ha-ha.

>>105731012
>The gap closes with high frequency lower latency ram
Yeah intlel goes from 30% behind to 29%
All you need is to pay more money.for top tier RAM, overclock it and pray that it's stable

>>105730963 (OP)
>pay extra for a slower cpu
I don't think I will

>>105730963 (OP)
intel doesnt suck. still not upgraded from my 10700 but have a 5070ti which is running at 165fps 1440p ultra and the cpu has the gpu core at 98%. intel cpu’s are badass. however imwill be going amd next for the length of time sockets are supported. sick of having to change mobo when buying a cpu.

>>105731669
i'm OP and i have a 10700k aswell, but every cpu intel made after that one basically sucks ass

>>105730963 (OP)
>>105731788
op, can you tell me what info they store on the cpu with the 98mb cache please? also, i have the non k but always get the 4.7 boost speed. its a great cpu but its getting warm now. peaked at 76c yesterday. but it was a very warm day. turned the gpu fans up to 47% and it helped cool it to 58c avg

>>105730980
amd's memory controller is actually great, the problem is the GMI links they use (aka infinity fabric links). they are both their greatest strength and their greatest weakness, on the one hand , GMI is very good for supporting chiplets cheaply which reduces the cost of manufacturing, on the other it makes it hard to get full throughput. you look at what intel is doing with arrow lake and its clear that amd's cheapo approach with cache is superior despite being much less complicated , it performs better most of the time.

>>105730963 (OP)
x3d chips are garbage, they only give you fps in games and they suck at anything outside gaming

>>105731590
Did you even read the chart you posted? You're so dumb it hurts to watch.

>>105731959
>they only excel at the thing people buy them for

>>105731960
The chart clearly shows your fancy x3d chip is slower than cpus half the price

>>105731590
This is such a dumb chart lol


to the uninitiated, larger caches have longer access times, however this test is measuring ram latency, so while the number is larger than the other chips, anything inside the cache is going to be accessed at like 13ns, and tripling the cache means that you don't have to go to main memory nearly as often, making this test metric completely worthless.

>>105732008
Cope. If cache is so good why are you using a lowly 96MB cache chip? broadwell i7 has 128MB of cache.

>>105732037
broadwell-C's cache was primarily for the graphics and it was not that fast, being able to use it as an l4 cache was a bonus feature.

The 96mb of l3 on zen4/5 operates at around 700gb/s and has a latency of 12-16ns

broadwell's l4 cache is 128mb operates at 50gb/s and has a latency of about 40ns.

>>105730963 (OP)
Intel did it back in 2015 with the 5775C. It was equivalent to a 10600k in games that utilised it.
Gamers aren't their main consumer base anyway. It's like 10-15% of their market.

>>105731590
Why is that so high? I'm running XMP on my 13700k with my 13700k and get 50ns.

>>105732423
the arrow lake cpus and zen cpus are both chiplet based, and take a latency hit due to the interconnect be tween dies, th everything intel , raptorlake to nehalem uses a monolithic die, and do not have a die to die interconnect.

>>105732573
Yeah I am aware but there are 14th gen on that list.

>>105732581
probably bad memory config, other reviews dont have it that high

>>105730963 (OP)
>When AMD couldn't profitably put more cache on the same 2D surface as the CPU cores, they went vertical.

Lmao you make it sound like AMD are the geniuses here. It was all TSMC. AMD just said oh shit we would like some of that.