• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Discussion Intel current and future Lakes & Rapids thread

Page 548 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

mikk

Diamond Member
May 15, 2012
3,253
1,065
136

5GHz reported.

Moat of my tests are done whilst I was testing GPU settings, I'll need to look for one where I actually ran the CPU test. Although my own RAM is far worse, only DDR4-3200 cl14 so not really comparable to the DDR4-3866 cl14 used on the 12900K.

Probably OC 5950X, this doesn't tell much. It's unlikely 3dmark would run at 4950 Mhz on a 5950x default because the core scaling isn't bad in 3dmark. Typical gaming clock speed is below 4950 Mhz.
 

uzzi38

Golden Member
Oct 16, 2019
1,817
3,654
116
Probably OC 5950X, this doesn't tell much. It's unlikely 3dmark would run at 4950 Mhz on a 5950x default because the core scaling isn't bad in 3dmark. Typical gaming clock speed is below 4950 Mhz.
Dude, I just explained what's going on. Read what I wrote up there. That score I linked is my own 5950X which is running PBO. When I run the test myself, sustained clocks are in the 4.55-4.6GHz range, but the validation screen shows 5GHz.
 

JoeRambo

Golden Member
Jun 13, 2013
1,246
1,183
136
Gotcha, thanks. In other words, we still just have to wait. Unless you're Joe Rambo and you have decided what the strengths and what the weaknesses of the new CPU are already.
It's not like one can go much further than 3866CL14, and frankly FH4 and Tombraider CPU benches is all i need to know about CPU performance in games. Looks like mobile chip memory subsystem that had to be repurposed to desktop and compromises were made.
 
May 1, 2020
70
81
51
This product and those of the next few years were most probably shaped under Jim Keller’s watch.
Well, IMHO Jim also had almost no impact. His stint was much shorter than originally planned (maybe due to internal differences) and he also was in a senior position.
 

geegee83

Junior Member
Jul 5, 2006
19
4
66
Well, IMHO Jim also had almost no impact. His stint was much shorter than originally planned (maybe due to internal differences) and he also was in a senior position.
Senior positions usually make the big decisions and set direction. Setting the wrong direction can set any company back by years.

But agree that deep technical details probably not.
 

inf64

Diamond Member
Mar 11, 2011
3,150
2,172
136
I'm not really impressed with these early gaming benchmarks :/. While Golden Cove appears to have ~10-11% higher *general int IPC* versus Zen3, it seems to be suffering from memory latency issues in games, even with very fast DDR4 (what will happen with vanilla DDR5?). If nothing changes, Zen3D 12 core part will likely be the dominant gaming chip for enthusiasts, easily. Intel is also running at very high clocks on GC so there is not much room left until they launch a refresh next year. But again, next year comes the big hammer Zen4 so even though it's great to see a super competitive intel again, it seems like they will still lag behind until Meter Lake comes.
 
  • Like
Reactions: Tlh97 and Joe NYC

lobz

Golden Member
Feb 10, 2017
1,818
2,353
136
It feels like the exact opposite is happening now here on this forum, than what happened with Rocket Lake, where people believed it's gonna rock(et) in gaming, then it flopped harder than I ever imagined. Please let's just wait for some people who actually know how to test gaming on CPUs.
 
  • Like
Reactions: Tlh97 and Markfw

Ajay

Diamond Member
Jan 8, 2001
9,485
3,951
136
I'm not really impressed with these early gaming benchmarks :/. While Golden Cove appears to have ~10-11% higher *general int IPC* versus Zen3, it seems to be suffering from memory latency issues in games, even with very fast DDR4 (what will happen with vanilla DDR5?). If nothing changes, Zen3D 12 core part will likely be the dominant gaming chip for enthusiasts, easily. Intel is also running at very high clocks on GC so there is not much room left until they launch a refresh next year. But again, next year comes the big hammer Zen4 so even though it's great to see a super competitive intel again, it seems like they will still lag behind until Meter Lake comes.
So what's you prediction for Raphael?
 

inf64

Diamond Member
Mar 11, 2011
3,150
2,172
136
So what's you prediction for Raphael?
I think Raphael with V Cache will be between 40 to 50% faster than vanilla Zen3 . That is a combination of 20-25% higher IPC, V cache effect(~15% in games) and higher boost and base clocks. That should be plenty enough to keep it above any refresh of Golden Cove/Atom, especially in games. Interesting things come with Meteor Lake Vs Zen5, as both have insane IPC targets.
 

Ajay

Diamond Member
Jan 8, 2001
9,485
3,951
136
I think Raphael with V Cache will be between 40 to 50% faster than vanilla Zen3 .
So the current thinking is Raphael be have V Cache by default? That will be pricey, more packaging costs and lower net chiplet yield. Seems like the cost effective solution will be vanilla Zen4 chiplet shipped to market first.
 

maddie

Diamond Member
Jul 18, 2010
3,513
2,497
136
So the current thinking is Raphael be have V Cache by default? That will be pricey, more packaging costs and lower net chiplet yield. Seems like the cost effective solution will be vanilla Zen4 chiplet shipped to market first.
Why?

Being designed for V-cache from the onset, what if there is no integral L3 cache but all is in the "V" chiplet? With different libraries used, we might have a reduction in total silicon area that negates the increased packaging costs.
 

pakotlar

Senior member
Aug 22, 2003
464
6
81

Hulk

Diamond Member
Oct 9, 1999
3,116
563
126
I have been out of gaming for quite a few years I admit. But when I was gaming it always seemed to be the case that the GPU was the bottleneck unless you had a really old CPU. From what I'm reading here that's not the case today right? Meaning something like a 9900K wouldn't be a good gaming CPU?

Lots of talk about CPU gaming prowess here. Just wondering how important this actually is for gamers as opposed to Intel/AMD marketing?
 

majord

Senior member
Jul 26, 2015
391
409
136
Yeah while this is great it’s an absolutely ridiculous score for a 5950x.

My PBO -15 to -30 curve optimizer 5950x on air gets 14.4K at best, typically 13.9k-14.1k

Alder Lake will also be overclockable.
As explained earlier that's to do with an issue with 32 threads in time spy. My pbo 5900x gets just shy of 14k
 
Last edited by a moderator:

pakotlar

Senior member
Aug 22, 2003
464
6
81
As explained earlier that's to do with an issue with 32 threads in time spy. My pbo 5900x gets just shy of 14k
This is not a typical score for 5900x, afaik.

12.6k stock, 13k with PBO, which echoes other stock scores I've seen.

Also I see nothing about a specific issue with 32 threads, just hyper threading in general. Alder Lake 12900K also has 8 hyper threads.
 

majord

Senior member
Jul 26, 2015
391
409
136
W
This is not a typical score for 5900x, afaik.

12.6k stock, 13k with PBO, which echoes other stock scores I've seen.

Also I see nothing about a specific issue with 32 threads, just hyper threading in general. Alder Lake 12900K also has 8 hyper threads.
*

Whatever the exact issue is , it's clearly worse the higher core/thread count , as evidenced by the Similar 5900x and 5950x scores on your link.

Just look at Firestrike ex physics results instead . 12900k showing Similar score to a 5950x from various other sources .
 

PJVol

Member
May 25, 2020
28
16
41
Why bother testing CPU in Firestrike or Timespy, when there is a CPU Profile benchmark for this purpose?
 

uzzi38

Golden Member
Oct 16, 2019
1,817
3,654
116
This is not a typical score for 5900x, afaik.

12.6k stock, 13k with PBO, which echoes other stock scores I've seen.

Also I see nothing about a specific issue with 32 threads, just hyper threading in general. Alder Lake 12900K also has 8 hyper threads.
It's an issue with high core count CPUs, not SMT/HT. Once you have more than 16 threads TS starts to struggle.

The question I have is how does Alder Lake react considering the fancy new scheduler and all. Does the scheduler actually assign TS to just the 8 big cores only or is TS running on all 24 threads?
 

arandomguy

Senior member
Sep 3, 2013
545
169
116
I have been out of gaming for quite a few years I admit. But when I was gaming it always seemed to be the case that the GPU was the bottleneck unless you had a really old CPU. From what I'm reading here that's not the case today right? Meaning something like a 9900K wouldn't be a good gaming CPU?

Lots of talk about CPU gaming prowess here. Just wondering how important this actually is for gamers as opposed to Intel/AMD marketing?
A lot of angles to consider here.

While the 9900k is somewhat old from a launch date perspective it isn't really from an actual practical perspective as Intel's per core design had effectively been stalled since Skylake in 2015 until Rocket Lake this year. Rocket Lake likely however had compromises and scale back due to the manufacturing process issues (even with Alder Lake I'd wonder if it does to some extent as well). AMD was really playing catch up still until Zen 3 and so it wasn't until Zen 3 that we've had what would really be considered a significant per core performance jump for gaming. Gaming did start to scale beyond 4c/8t but not much beyond that, and since the 9900k is a 8c/16t it's fine in that respect. This means in practice despite the age the 9900k is really just 1 tier down from Zen 3 CPUs for gaming.

We're still in the transitional phase for the console generations. There is much more limited scaling on the CPU side for gaming compared to graphics. It'll be interesting to see if once we fully transition to what extent games start to leverage the large CPU capability jump this generation and how that impacts the PC side when the CPU gap is nowhere near what it enjoyed over the last gen.

PC gaming is really broad nowadays and the measurables have changed. For instance I know people that were interested when it was noticed Zen 3 was showing a high consistency of running Overwatch well above 300 FPS even more so than Intel CPUs. Now does something like that matter for the "graphics review games" segment that just wants 60 fps at higher resolutions? Different segments. Higher refresh at 120hz+ now is still now considered high refresh anymore for gaming. There are also games that have struggled to even hit 60 fps due to the CPU historically in segments (eg. notably Fallout 4, which I'll mention a bit below).

This is a bit of a side bar issue I have but I don't feel most reviewers really even look at CPUs for gaming in a relevant manner because they effectively look at it from the same perspective as a graphic's card review but with CPUs instead. The Fallout 4 benchmark thread on these forums awhile back for example was a much more relevant from a CPU gaming performance stand point than Anandtech's actual CPU reviews.

Related to the above I also feel gaming numbers have some proxy for non gaming and productivity as well. I have a problem with how most reviewers tackle productivity as it just simplifies productivity to throughput based tests essentially. I do 3D rendering for example, but the thing is I use GPUs now for the actual throughput part of rendering (CPUs just don't compare now). Which means for me a better CPU for rendering isn't the one that actually does those render tests faster but which is faster in the actual creation portion (which tend to somewhat almost align more with gaming). But most reviews don't look at productivity from that angle.

From an upgradeability stand point my experience is also that for most users the GPU is the most upgraded. While there is often vocal talk about socket/mobo support for future upgrades for CPUs it's not really taken advantages of to anywhere near the same extent. Then of course monitors (displays) and display resolution also see rather long life spans (noticing a significant amount of people upgrading now to 120hz+ versus resolution increases). So while you may not have RTX 3090 or 6900 XT now to pair with your CPU, you may eventually be using a GPU that well eclipses those. I wouldn't be surprised if a significant amount of early 2010 CPU buyers for example eventually used GPUs several times faster than the top end GPU available at the time with the same CPU.
 
Last edited:

lobz

Golden Member
Feb 10, 2017
1,818
2,353
136
It's an issue with high core count CPUs, not SMT/HT. Once you have more than 16 threads TS starts to struggle.

The question I have is how does Alder Lake react considering the fancy new scheduler and all. Does the scheduler actually assign TS to just the 8 big cores only or is TS running on all 24 threads?
That's exactly why I thought we were still not very much closer to 'the truth' 🙂😁
 

itsmydamnation

Platinum Member
Feb 6, 2011
2,249
1,925
136
Err scheduler schedules it doesn't pick number of threads the app does that itself. So Its safe to assume 1 software thread per hardware thread regardless of big little magic. At that point if it does anything other Then assign those across all hardware threads then the Scheduler is a complete failure.
 
  • Like
Reactions: coercitiv

Zucker2k

Golden Member
Feb 15, 2006
1,528
867
136
Err scheduler schedules it doesn't pick number of threads the app does that itself. So Its safe to assume 1 software thread per hardware thread regardless of big little magic. At that point if it does anything other Then assign those across all hardware threads then the Scheduler is a complete failure.
I saw a post on another forum where these ADL numbers were better compared to average 5950x numbers from the UL website iirc. What's causing confusion here is a couple of members posting their own jacked-to-the-gills systems numbers in comparison to a sample of one ADL running on who knows what? Reviews should paint a clearer picture because hopefully no reviewer would compare tweaked systems vs stock.
 

uzzi38

Golden Member
Oct 16, 2019
1,817
3,654
116
I saw a post on another forum where these ADL numbers were better compared to average 5950x numbers from the UL website iirc. What's causing confusion here is a couple of members posting their own jacked-to-the-gills systems numbers in comparison to a sample of one ADL running on who knows what? Reviews should paint a clearer picture because hopefully no reviewer would compare tweaked systems vs stock.
They said that their Alder Lake system was running DDR4-3866cl14 though? Like it's right there in the same post. Even on it's own that's already going to give you a far better result than your average 5950X system would.
 

DrMrLordX

Lifer
Apr 27, 2000
17,843
6,819
136
Why?

Being designed for V-cache from the onset, what if there is no integral L3 cache but all is in the "V" chiplet? With different libraries used, we might have a reduction in total silicon area that negates the increased packaging costs.
Raphael will probably use Genoa dice.
 
  • Like
Reactions: Tlh97 and Ajay

LightningZ71

Senior member
Mar 10, 2017
942
908
136
So, are we now demanding that reviewers only use jedec spec ddr4-3200 ram again? That's the way OEM machines will come locked to after all. It is the "official" limit by both vendors.
 
  • Haha
Reactions: Zucker2k

ASK THE COMMUNITY