Discussion Intel Nova Lake in H2-2026: Discussion Threads

[msj10]

On the Topic of NVL Arctic Wolf vs Coyote Cove which will have higher perf/clock what do you guys think?

my guess is that both will get around 15% IPC improvement so the gap will stay the same

 

[511]

My expectations are 14-19% for P core and 25% for the E core team based on track record for last 5 years.

 

[Kepler_L2]

My expectations are 14-19% for P core and 25% for the E core team based on track record for last 5 years.

Since the E-core team has to spend much of their transistor count boost on AVX10/APX I don't think their IPC gain will be over 20%.

 

[511]

Since the E-core team has to spend much of their transistor count boost on AVX10/APX I don't think their IPC gain will be over 20%.

From what i know it is going to be Zen4 like AVX-512 implementation and raichu said 20% over Darkmont.

 

[gdansk]

Are you conflating IPC and performance? Mont has the easy path of pushing freq yet, so expect that to be some of the 20%.

 

[511]

Are you conflating IPC and performance? Mont has the easy path of pushing freq yet, so expect that to be some of the 20%.

I meant perf/clock perf is easy they can do 5Ghz easily tbh with N2/N2P

 

[gdansk]

I meant perf/clock perf is easy they can do 5Ghz easily tbh with N2/N2P

So, maybe I misread. What's the IPC gain and also the overall performance gain you expect for next Mont?

 

[511]

So, maybe I misread. What's the IPC gain and also the overall performance gain you expect for next Mont?

20-25% ish overall 28-32% peak perf improvements i am assuming 5Ghz for the next mont lol

 

[gdansk]

20-25% ish overall 28-32% peak perf improvements i am assuming 5Ghz for the next mont lol

Hmm, well it seems to have escalated from 20% perf somehow.

 

[511]

Hmm, well it seems to have escalated from 20% perf somehow.

I always meant 20% minimum IPC gains lol and 8% from frequency is not far fetched aka 5 Ghz

 

[gdansk]

I always meant 20% minimum IPC gains lol and 8% from frequency is not far fetched aka 5 Ghz

So if I'm extrapolating right, you expect it to be as good as an M3 Pro P core in SPECint rate-1?

 

[dullard]

I always meant 20% minimum IPC gains lol and 8% from frequency is not far fetched aka 5 Ghz

Single threaded: 5 GHz Monts and maybe 6 GHz Cove is possible but I'm not expecting that much.

But, since most Nova Lake CPUs get double the core count without nearly that much change in TDP, the all core multi-threaded frequencies probably won't go up.

 

[511]

So if I'm extrapolating right, you expect it to be as good as an M3 Pro P core in SPECint rate-1?

slightly behind

 

[511]

But, since most Nova Lake CPUs get double the core count without nearly that much change in TDP, the all core multi-threaded frequencies probably won't go up.

for the 24C variants i was hoping the max turbo for all cores to go up as well though not by much

 

[Fjodor2001]

Single threaded: 5 GHz Monts and maybe 6 GHz Cove is possible but I'm not expecting that much.

But, since most Nova Lake CPUs get double the core count without nearly that much change in TDP, the all core multi-threaded frequencies probably won't go up.

Yeah, what clocks can be expected when all 52C are pumping max MT workloads simultaneously?

 

[dullard]

for the 24C variants i was hoping the max turbo for all cores to go up as well though not by much

So you are talking about the Ultra 5 version of Nova Lake (you know the 24 core version of Nova Lake). Otherwise you'd be comparing the Ultra 9 Arrow Lake to the Ultra 5 Nova Lake which is just a pretty unusual comparison to make on multiple levels (price, power, bin level, intended market, etc).

The Ultra 5 Arrow Lake has 4.9 GHz to 5.2 GHz P cores and 4.5 GHz to 5.6 GHz E cores. How much do you think these speeds will go up for the Ultra 5 Nova Lake?

 

[511]

So you are talking about the Ultra 5 version (you know the 24 core version of Nova Lake).

Yes lets forget about 48C for comparison sake cause that would be power limited

The Ultra 5 Arrow Lake has 4.9 GHz to 5.2 GHz P cores and 4.5 GHz to 5.6 GHz E cores. How much do you think these speeds will go up?

typo but maybe 8-10% frequency increase

 

[Fjodor2001]

Yes lets forget about 48C for comparison sake cause that would be power limited

What if we do not forget about the 48/52C case, any estimate for clocks then?

We'll be going from ARL-S 8+16 (=24) @ 125W TDP, to NVL-S 16+32+4 (=52) @ 150W TDP. So a lot less power per core. That said, we'll also be going from TSMC N3B to N2, and the last few W per core does usually not bring that much extra perf.

 

[511]

What if we do not forget about the 48/52C case, any estimate for clocks then?

Nope cause don't even have a baseline to estimate have to run ARL-S at PL1=PL2 at 125W to get estimate as PL2=250W

 

[Fjodor2001]

Nope cause don't even have a baseline to estimate have to run ARL-S at PL1=PL2 at 125W to get estimate as PL2=250W

What about the graph in this post:

Page 40 - Discussion - Intel Nova Lake in H2-2026: Discussion Threads

Page 40 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Could it perhaps give some clues to be used for a guesstimate?

 

[511]

Could it perhaps give some clues to be used for a guesstimate?

it's not a V/F Curve still difficult i would be throwing number around 🤣 🤣

 

[Fjodor2001]

Some benchmark leaks would be nice. But I guess we won't get that for another 6-9 months or so, assuming NVL-S will be released in 2026Q4.

 

[511]

PTL is scheduled like October November and we haven't gotten any leaks lol

 

[poke01]

PTL is scheduled like October November and we haven't gotten any leaks lol

There is an event today. Info should come out soon.

 

[511]

There is an event today. Info should come out soon.

That is NDA event