Discussion Intel Nova Lake in H2-2026: Discussion Threads

[Fjodor2001]

In Cinebench. Why extrapolate, repeatedly, from one single benchmark chart that doesn't represent your workload?

If you expand the benchmark scope what happens? Where does Arrow Lake fall apart? Where does scaling fall apart in general? Not in Cinebench.

That’s the only power scaling test result for MT workloads with these two CPUs that I could find. But if you have other similar ones, please share. Could be interesting to compare.

 

[511]

Someone needs to run spec n copy

 

[adroc_thurston]

Someone needs to run spec n copy

chinese guy (forgot his blog) tortured 682 ARL-H and STX1 in SIR2017.
The results were funny.

 

[511]

chinese guy (forgot his blog) tortured 682 ARL-H and STX1 in SIR2017.
The results were funny.

Link Please
i think you meant david huang

 

[adroc_thurston]

i think you meant david huang

yessir

TSMC救不了Intel：酷睿Ultra 7 255H（Arrow Lake H45）测试 | David Huang's Blog

 

[gdansk]

Link Please
i think you meant david huang

TSMC救不了Intel：酷睿Ultra 7 255H（Arrow Lake H45）测试 | David Huang's Blog

Don't skip power consumption, efficiency sections.

 

[adroc_thurston]

TSMC救不了Intel：酷睿Ultra 7 255H（Arrow Lake H45）测试 | David Huang's Blog

Don't skip power consumption, efficiency sections.

too late, I've already poasted the chart.

 

[511]

I did read that blog. SPEC is sensitive to cache and ARLs L3 has been horrendously bad.

Even 4 Core Skymont beats 4C LNC shame to the P cores Without Skymont ARL would have been dud.

 

[adroc_thurston]

I did read that blog. SPEC is sensitive to cache and ARLs L3 has been horrendously bad.

The interesting part is 8c/8t efficiency tie between Z5c and SKT despite node delta and SKT having like 3 times the accessible LLC.

 

[511]

The interesting part is 8c/8t efficiency tie between Z5c and SKT despite node delta and SKT having like 3 times the accessible LLC.

Yeah but the Cache latency is not the same as for node Delta it's N3B it's not a full node better than N4P at best single digit difference.

 

[gdansk]

ARLs L3 has been horrendously bad.

And NVL-S x2 cache & coherency will have to do so much more.
Hope they got their top minds on it.

 

[adroc_thurston]

it's N3B it's not a full node better than N4P at best single digit difference.

oh no it is much better at low power.

 

[511]

oh no it is much better at low power.

like how much ~10% Intel could have spammed more skymont tbh to spam MT like they are doing with Clearwater Forest too bad it is launching this late should have been able to hold off Turin Dense.

 

[adroc_thurston]

like how much ~10%

20-ish %.

 

[511]

20-ish %.

That's a lot I wonder how much is the botched design is removing
Cause LNL Curve vs ARL curve is like 50% difference for P core.

 

[Darkmont]

And NVL-S x2 cache & coherency will have to do so much more.
Hope they got their top minds on it.

Their fabric/LLC top minds™ have been on a generational run of shatting the bed. Only Allah knows how bad NVL-S bLLC latency will be.

 

[511]

Their fabric/LLC top minds™ have been on a generational run of shatting the bed. Only Allah knows how bad NVL-S bLLC latency will be.

I wonder why they don't have been laid off they need to re organize this team they have been s**ting the bad for so long.
Clearwater Forest has been bottlenecked due to Fabric/L3.

 

[adroc_thurston]

That's a lot I wonder how much is the botched design is removing

plenty.
LNC physdes has been a shitfest and you know it.

Only Allah knows how bad NVL-S bLLC latency will be.

can we has been beat GNR and it's mighty 130 cycles near LLC latency?

 

[gdansk]

Their fabric/LLC top minds™ have been on a generational run of shatting the bed. Only Allah knows how bad NVL-S bLLC latency will be.

I'm actually hopeful in this respect. If Big Cache technology didn't help a fair deal they'd probably have canned it. But it made it through.

 

[511]

plenty.
LNC physdes has been a shitfest and you know it.

I Know they fixed most of it on LNC on LNL.

can we has been beat GNR and it's mighty 130 cycles near LLC latency?

Even Desktop Ram has better latency.

 

[511]

I'm actually hopeful in this respect. If Big Cache technology didn't help a fair deal they'd probably have canned it. But it made it through.

At the same latency at ARL but 4X the cache is still better but the botched L3s have been a issue their cores are not getting fed properly.

 

[Darkmont]

Even Desktop Ram has better latency.

Woah woah woah there buddy it's not THAT bad Christ have mercy.

 

[adroc_thurston]

Even Desktop Ram has better latency.

not ns. cycles.

 

[511]

Woah woah woah there buddy it's not THAT bad Christ have mercy.

not ns. cycles.

My bad lol i got it in the heat of moment how their L3 has been sucking ass

 

[igor_kavinski]

Even Desktop Ram has better latency.

Well, it's a good time for Intel to re-introduce its eDRAM (at least 512MB) on cheap non-K PCs with JEDEC RAM. That could make them viable gaming machines.