• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Intel Nova Lake in H2-2026: Discussion Threads

Page 67 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
adroc_thurston said:
he point is that you're going from 1*110mm^2 N3 to 2*150mm^2 N2p on the very top.
You WILL pay more because that's a lot more of more expensive Si.
Click to expand...
No disagreement regarding the pricing of the 2x24C bLLC SKU. It's going to rightfully be in the price range of the HEDT platform of old.

The 24C bLLC part definitely does have justification for a price increase, but I'd guess it to be more on the order of going from the current $600 MSRP of the ultra 9 285k to $700.

The more pertinent question from the enthusiast point of view looking for the best price/performance ratio is whether there's going to be an ultra 7 bLLC variant any time soon.
 
Given how expensive the boards are going to have to be (if it's going to be a 400+ W monster), kinda wondering why it isn't quad (eight) channel and aimed at HEDT.
 
jpiniero said:
Given how expensive the boards are going to have to be (if it's going to be a 400+ W monster), kinda wondering why it isn't quad (eight) channel and aimed at HEDT.
Click to expand...
W980 board, which is what I'd want, will probably be insane. My W880 board was $450 already and it was the cheapest I could find.
But at least I'll be able to reuse my 48GB DDR5 ECC UDIMMs.
 
jpiniero said:
Given how expensive the boards are going to have to be (if it's going to be a 400+ W monster), kinda wondering why it isn't quad (eight) channel and aimed at HEDT.
Click to expand...
Because not all MT workloads need that. And because faster supported DDR5 speeds. And because likely improved caches. And because much cheaper than HEDT, effectively killing off the lower part of the HEDT range.

Also, I wonder if that amount of power really has to be supported on all motherboards. And whether it’s release specs, or early development specs.
 
Fjodor2001 said:
Because not all MT workloads need that. And because faster supported DDR5 speeds. And because likely improved caches. And because much cheaper than HEDT, effectively killing off the lower part of the HEDT range.

Also, I wonder if that amount of power really has to be supported on all motherboards. And whether it’s release specs, or early development specs.
Click to expand...
what kind of MT workloads do you do on your desktop?
 
Fjodor2001 said:
What about the 2x24C without bLLC, what $ amount do you put on that and based on what?
Click to expand...
That's one which I'm not certain they'll end up making? Is there really adequate differentiation between 2x24C bLLC and 2x24C standard? Are they going to lose sales to the AMD alternative due to not offering it? Or will whatever users who want 2x24C not care and pay the bLLC premium even if it doesn't benefit their workload?

I suspect that the 2x24C bLLC part is going to be a very low volume 'halo' part whose primary goal is to win all the benchmarks and drive sales of the lower tier parts.
 
Khato said:
That's one which I'm not certain they'll end up making? Is there really adequate differentiation between 2x24C bLLC and 2x24C standard?
Click to expand...
At least in earlier leaks I’ve seen NVL-S SKUs both with and without bLLC.

I guess it’s the same as with the Zen X3D models. The bLLC SKUs will cost more, so they want to provide SKUs without bLLC too for those that do not think it’s worth the extra cost.
 
Fjodor2001 said:
And you're still at it with your usual stuff. If you have something to counter, share.
Click to expand...
You've already been corrected on why 1.2 would be wrong. Follow the replies to your comments. Learn as you read the threads. Listen. You might be taken more seriously, I hope you do because you're an enthusiastic poster just, struggle to follow a bit. Look at the die yield calculator
 
Fjodor2001 said:
Except that your crystal ball has a poor track record and has been proven wrong so many times in the past.

But please go ahead and pin it to all threads in the CPU forum, so you don't have to repeat it over and over again in all threads.
Click to expand...
While a fair comment itself. It's an interesting choice when one is standing in the most fragile of glass houses
 
inquiss said:
You've already been corrected on why 1.2 would be wrong. Follow the replies to your comments. Learn as you read the threads. Listen. You might be taken more seriously, I hope you do because you're an enthusiastic poster just, struggle to follow a bit. Look at the die yield calculator
Click to expand...
It was just a rough estimate and there was no conclusion on what number should be used instead of 1.2. There were just different aspects being mentioned by others, which could affect it somewhat up or down.

But if you know what better and more accurate number should be used instead of 1.2, then please share.
 
Last edited:
Fjodor2001 said:
I guess it’s the same as with the Zen X3D models. The bLLC SKUs will cost more, so they want to provide SKUs without bLLC too for those that do not think it’s worth the extra cost.
Click to expand...
Keep in mind that Intel's approach is not the same as AMD's X3D models. I don't recall if the X3D used the same packaging as the normal chiplets or not? Pretty sure they do, but even if not that would just require two variants of the FCBGA packaging.

By comparison the bLLC die require a different Foveros base die. So for Intel the question is whether there's enough volume/demand to justify both a 2x24C bLLC base die and a standard 2x24C base die. I don't see the demand being there... but that's in part because I remain unconvinced of the need for even 24 cores on mainstream desktop platforms. (Yes, I know some fraction of a percent of users do make proper use of the higher core counts, but in general it's just wasted silicon.)
 
Khato said:
Keep in mind that Intel's approach is not the same as AMD's X3D models. I don't recall if the X3D used the same packaging as the normal chiplets or not? Pretty sure they do, but even if not that would just require two variants of the FCBGA packaging.

By comparison the bLLC die require a different Foveros base die. So for Intel the question is whether there's enough volume/demand to justify both a 2x24C bLLC base die and a standard 2x24C base die. I don't see the demand being there... but that's in part because I remain unconvinced of the need for even 24 cores on mainstream desktop platforms. (Yes, I know some fraction of a percent of users do make proper use of the higher core counts, but in general it's just wasted silicon.)
Click to expand...
Yes bandwidth to core ratio's be horrible. There are so many constraints for this thing to be performant.

More cache always helps but it helps less and less the more you have. I just don't see what the workloads tare hat it's going to be a massive bandwidth amplifier are.
 
itsmydamnation said:
Yes bandwidth to core ratio's be horrible. There are so many constraints for this thing to be performant.

More cache always helps but it helps less and less the more you have. I just don't see what the workloads tare hat it's going to be a massive bandwidth amplifier are.
Click to expand...
Should be similar for 48T Zen6. If anything NVL-S has the benefit of likely higher DDR5 speeds.
 
poke01 said:
no they don't, all Intel needs to do to win is to beat the 10800X3D. thats it.

Every other SKU is low volume
Click to expand...
Not true. It may be the best selling SKU, but not the only one that sells.

Anyway, the situation is the same for AMD. And both have such top 48T / 48C SKUs in their lineup with dual X3D / bLLC.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top