Discussion Intel current and future Lakes & Rapids thread

[uzzi38]

Turning off SMT in Timespy gives you a 2-3k higher cpu score when running a 5950x
(all the others 3dmarks don't punish you for having 32 threads, so SMT can be enabled when on running those)

I take it back, this is actually a thing.

Huh. The more you know I guess.

 

[mikk]

I scored 6 081 in Port Royal

AMD Ryzen 9 5950X, AMD Radeon RX 6700 XT x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

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]

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]

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.

 

[BorisTheBlade82]

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]

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]

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.

 

[lobz]

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.

 

[Ajay]

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]

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]

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]

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]

Sophia Script slimmed Windows10, what does it matter ?
View attachment 51432 View attachment 51433 View attachment 51434 View attachment 51435 View attachment 51436

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.


Sent from my iPhone using Tapatalk

 

[Hulk]

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]

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

 

[pakotlar]

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.

AMD Ryzen 9 5950X and 5900X Review: Zen 3 Breaks the 5 GHz Barrier

AMD's Magnum Opus

www.tomshardware.com

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]

W

This is not a typical score for 5900x, afaik.

AMD Ryzen 9 5950X and 5900X Review: Zen 3 Breaks the 5 GHz Barrier

AMD's Magnum Opus

www.tomshardware.com

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]

Why bother testing CPU in Firestrike or Timespy, when there is a CPU Profile benchmark for this purpose?

 

[uzzi38]

This is not a typical score for 5900x, afaik.

AMD Ryzen 9 5950X and 5900X Review: Zen 3 Breaks the 5 GHz Barrier

AMD's Magnum Opus

www.tomshardware.com

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]

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.

 

[lobz]

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]

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.

 

[Zucker2k]

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]

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]

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.