Is it really over for Intel®?

[koohansen]

ah, was planning on upgrading to this.
still on a 10th gen..

15th gen coming, and Z890 just released
wait a while

 

[R6exR6]

Question:
Will 9800X3D still be good enough for 60 fps @4K max RT settings in 10 years' time (assuming GPU not bottleneck)?

Fyi, my i5-3570K can't even do 30 fps only like 5 years later into AAA gaming.

 

[iceblendedchoc]

Question:
Will 9800X3D still be good enough for 60 fps @4K max RT settings in 10 years' time (assuming GPU not bottleneck)?

Fyi, my i5-3570K can't even do 30 fps only like 5 years later into AAA gaming.

10 years maybe not. 5 to 6 years will be more likely.

the bar for max settings gaming with highest resolution just keep getting higher.

 

[SkyShroud]

Intel is targetting production, the integrated NPU made their intentions obvious. I am sure they know the latency is horrible for gaming, they totally gave up on this generation of gaming market. I think only next generation will they able to fix their tiles latency, for them, selling this mid tiles design is just to recoup their investments because they cannot let their TSMC nodes sit there doing nothing.

 

[kimsix]

local 285K prices about $50 cheaper than 9950X
Asus x870e and z890 mobo around same price
MSI and Gigabyte z890 is more ex than their x870e by $150

cudimm not here yet

arrow lake overclocking parameters like a lot, very time consuming

zen tuning seems so much lesser and easier for noobs

not appealing to DIY a CU 200 series ... for now

 

[Phen8210]

How many users really care about gaming other than the few gaming enthusiasts and few review sites?

Desktop does not even matter much in the personal computer market, let alone gaming PC.

Intel has problems, but not because of gaming performance at all.

PS: I am OS neutral and Intel/AMD/Apple neutral. I am using Mac Mini M1, Ryzen 5 5600G desktop (dual boot Windows/Linux) and Acer Swift 3 laptop with Intel Core i5-1135G7 CPU (dual boot Windows and Linux). I also bought two Intel Core Ultra 5 125H based Acer Swift 14 Go OLED laptops this year for my wife and daughter.

Most likely I will upgrade my laptop next year, and maybe Mac Mini as well probably

I think you are not thinking outside the box. Gaming PCs can be used for many things besides gaming. They are also ideal for video editing, 3D rendering, and IT experts who need good performance.

My laptop is actually better spec-ed than my desktop, benchmarks higher, and all that stuff, but when it comes to snappiness and responsiveness, it's nowhere close. There are tasks where my desktop always responds instantly compared to having to wait 2-3 seconds.

As for intel they are facing reputation damage, even if they made better CPU, serious users will still think twice before considering their product.

 

[m0n0n0ke]

 

[NightRaven49]

I wonder if intel is considering something similar (ie large amount of cache) to amd's vcache. That seems to be the logical direction to head towards to rather than trying to win clock wars which has increasing diminishing returns - more power used/heat for less gains, at least in the context of gaming.

Question:
Will 9800X3D still be good enough for 60 fps @4K max RT settings in 10 years' time (assuming GPU not bottleneck)?

Fyi, my i5-3570K can't even do 30 fps only like 5 years later into AAA gaming.

4k rt is generally a gpu bottleneck scenario, so any modern cpu should be good for many years to come. for example if u look at tpu's 285k 4k rt testing here: https://www.techpowerup.com/review/intel-core-ultra-9-285k/21.html#gaming-with-ray-tracing
the 5950x launched almost 5 years ago now, and still delivers highly playable performance today, mostly matching the latest cpus. u can also get a glimpse of how well the 78x3d may age by looking at the 58x3d results, which shares the same fundamental architecture as the 5950x but loses less than 10fps in the worst case compared to the 78x3d, even despite the sizeable performance uplift from zen3 to zen4.

 

[uselessbum]

Thank you for linking that. It's different from what I had in mind. My fault, I should have been more specific by referring to intel creating an architecture that can better make use of a larger cache.

The lack of impact with bigger cache in that test is due to the architecture of the 14th gen relying on clock speed and core counts, not cache. I'm sure it is about the same for AMD's non-x3d CPUs. Moreover, the difference in cache between the tested CPUs aren't as large as that of a X3D. Even steve mentioned at the end of the video that 14th gen does not make as much use of cache unlike the 10th gen where differences in performance was mainly attributed to cache size differences when set to the same frequency.

Perhaps a closer comparison would be that of intel's 5th gen broadwell. It still holds up quite well compared to its peers multiple generations later, thanks to its 128mb of L4 cache. That was an architecture designed to make use of a bigger cache. The closest thing intel had to a "x3d".

https://www.anandtech.com/show/1619...ive-review-in-2020-is-edram-still-worth-it/18

In other words, I think there is more performance (in the context of gaming) to be extracted from having an architecture that is specifically designed around a large cache than one that primarily relies on ever increasing core count and clock speeds. At least until game engines can make better use of higher core count and clock speeds.

 

[m0n0n0ke]

 

[watzup_ken]

How many users really care about gaming other than the few gaming enthusiasts and few review sites?

Desktop does not even matter much in the personal computer market, let alone gaming PC.

Intel has problems, but not because of gaming performance at all.

PS: I am OS neutral and Intel/AMD/Apple neutral. I am using Mac Mini M1, Ryzen 5 5600G desktop (dual boot Windows/Linux) and Acer Swift 3 laptop with Intel Core i5-1135G7 CPU (dual boot Windows and Linux). I also bought two Intel Core Ultra 5 125H based Acer Swift 14 Go OLED laptops this year for my wife and daughter.

Most likely I will upgrade my laptop next year, and maybe Mac Mini as well probably

I think the problem for Intel Arrow Lake is that the performance gain is not consistent across different workloads. I agree that CPUs are not always for gaming, but I won't want to pay more for something that does not consistently do well when there are cheaper alternatives. Power efficiency is important, but on a desktop, the higher power requirement is less of a problem, especially when it is not a challenge to cool the CPU.

 

[watzup_ken]

local 285K prices about $50 cheaper than 9950X
Asus x870e and z890 mobo around same price
MSI and Gigabyte z890 is more ex than their x870e by $150

cudimm not here yet

arrow lake overclocking parameters like a lot, very time consuming

zen tuning seems so much lesser and easier for noobs

not appealing to DIY a CU 200 series ... for now

For Intel Arrow Lake, you will need to run faster memory to make the most of it. So while everything seems fairly even in cost, you also need to factor in the higher cost due to the faster memory. And assuming about the same cost, I find it very difficult to recommend a product that does don't perform consistently better than what it is replacing. Unless you have a very specific use where Arrow Lake shines.

 

[Phen8210]

For Intel Arrow Lake, you will need to run faster memory to make the most of it. So while everything seems fairly even in cost, you also need to factor in the higher cost due to the faster memory. And assuming about the same cost, I find it very difficult to recommend a product that does don't perform consistently better than what it is replacing. Unless you have a very specific use where Arrow Lake shines.

As long as it's better than the 12th gen, it's an improvement already because the 13th and 14th gen people will shun it.

Overall, it is still an improvement from intel. Most important is that these series don't possess huge issues that occurred like their 13th and 14th gen, otherwise it will be the end for them.

 

[elmariachi]

For Intel Arrow Lake, you will need to run faster memory to make the most of it. So while everything seems fairly even in cost, you also need to factor in the higher cost due to the faster memory. And assuming about the same cost, I find it very difficult to recommend a product that does don't perform consistently better than what it is replacing. Unless you have a very specific use where Arrow Lake shines.

Running faster memory on Arrow Lake will not compensate for the loss of performance. It will narrow the gap but on an equivalent 13th/14th gen, latency is still behind by almost 20ns handicapped mainly by the ringbus at 3900Mhz even with tuned memory. That too with the stupid tile system layout.

Arrow Lake will be good for content creators but not for much else. Also, CUDIMM only works on Z890. Using it on Z790 will simply bypass the CKD module. Arrow Lake will this time target a completely different group of users.

 

[NightRaven49]

I should have been more specific by referring to intel creating an architecture that can better make use of a larger cache.

The lack of impact with bigger cache in that test is due to the architecture of the 14th gen relying on clock speed and core counts, not cache. I'm sure it is about the same for AMD's non-x3d CPUs. Moreover, the difference in cache between the tested CPUs aren't as large as that of a X3D. Even steve mentioned at the end of the video that 14th gen does not make as much use of cache unlike the 10th gen where differences in performance was mainly attributed to cache size differences when set to the same frequency.

i dont think cpu manufacturers design architectures based on what their competitors' architectures benefit from. its mostly a game of refining their previous architectures and learning from their design inefficiencies. for amd, one of the reasons why vcache provides such a monumental uplift is because of its chiplet design, which adds latency when doing memory access compared to the monolithic designs of intel since the beginning of time, which is also why it is important to get fast tight memory for ryzen. by adding cache directly on top of the cpu cores, it allows the cores to rely less on system memory. this was also basically discovered by accident with engineers trying out different workloads on defect epyc chips, rather than something that amd specifically engineered for. with arrow lake now moving to a chiplet like approach, maybe we can see glimpses of cache being more relevant, with memory speeds being an area of focus to boost performance where it once wasnt.

 

[watzup_ken]

Running faster memory on Arrow Lake will not compensate for the loss of performance. It will narrow the gap but on an equivalent 13th/14th gen, latency is still behind by almost 20ns handicapped mainly by the ringbus at 3900Mhz even with tuned memory. That too with the stupid tile system layout.

Arrow Lake will be good for content creators but not for much else. Also, CUDIMM only works on Z890. Using it on Z790 will simply bypass the CKD module. Arrow Lake will this time target a completely different group of users.

Agree, and that is why I wanted to bring up the point that we can't just buy any DDR5 to use with Arrow Lake if one is concerned about performance. I think its clear that it will need very fast DDR5 to get the most of it.

I feel Intel being quite new to chiplet design will figure something out in the longer run. For now, Arrow Lake and likely the next gen will be more like testing and refinement for future chiplet designs.

 

[watzup_ken]

As long as it's better than the 12th gen, it's an improvement already because the 13th and 14th gen people will shun it.

Overall, it is still an improvement from intel. Most important is that these series don't possess huge issues that occurred like their 13th and 14th gen, otherwise it will be the end for them.

The reviews I saw did not include Alder Lake, but looking at some loads where Alder Lake is right at the bottom of the charts, I suspect it does not consistently beat Alder Lake. And the worst part is, Alder Lake is actually fairly efficient because it is the first Intel 10nm chip that is not pushed to the limits like Raptor Lake.

 

[elmariachi]

Agree, and that is why I wanted to bring up the point that we can't just buy any DDR5 to use with Arrow Lake if one is concerned about performance. I think its clear that it will need very fast DDR5 to get the most of it.

I feel Intel being quite new to chiplet design will figure something out in the longer run. For now, Arrow Lake and likely the next gen will be more like testing and refinement for future chiplet designs.

Agreed as well. But running very fast DDR5 past 9400 will also push you into gear 4 which halfs controller frequency further from Gear 2 which would make it run even slower unless maybe you hit DDR5 15000mhz then perhaps you can see some performance. Like you said, it all seems like testing and refinement for future chiplet designs but memory on Arrow Lake at least for now no longer scales like it used to on the previous generations. Alot will have to come from refinement of the cpu architecture if Intel chooses to stay on chiplet designs.

 

[Phen8210]

The reviews I saw did not include Alder Lake, but looking at some loads where Alder Lake is right at the bottom of the charts, I suspect it does not consistently beat Alder Lake. And the worst part is, Alder Lake is actually fairly efficient because it is the first Intel 10nm chip that is not pushed to the limits like Raptor Lake.

Agree, and that is why I wanted to bring up the point that we can't just buy any DDR5 to use with Arrow Lake if one is concerned about performance. I think its clear that it will need very fast DDR5 to get the most of it.

I feel Intel being quite new to chiplet design will figure something out in the longer run. For now, Arrow Lake and likely the next gen will be more like testing and refinement for future chiplet designs.

Hmm, good points.. lets just call it error lake then.

 

[watzup_ken]

Hmm, good points.. lets just call it error lake then.

May be not error lake, but trial and error lake? Intel's gotta start somewhere. I feel Intel was rapidly losing market share because of the immense power requirement from their previous CPUs. Which is why ARM is gaining popularity in the mobile and data center processors.