• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Intel’s Nova Lake-S LGA1954 52 cores

In terms of gaming performance, I was cautiously optimistic for Arrowlake-S and seeing how that turned out, I'm not even optimistic anymore for Novalake-S.

Reminds me of how I was so optimistic for Bulldozer and after that launched, I bought Sandybridge and never looked back for the next 11 years.
Its really not that bad. With the fixes in place and 200s boost implemented, arrow lake is crazy good value.
 
Are Intel CPUs bottlenecked as much by memory latency? That's the main thing a huge cache helps. Last I checked the memory throughput/latency was a huge limitation on Zen, which is why when you give it a big fuck-off cache suddenly it's a superchip in real time tasks.
Would Intel's arch experience the same boost?
Intel already had CPUs with large cache. Broadwell with 128MB L4 cache in Broadwell and it @4.2GHz often beat latter Skylake CPUs 5Ghz in games.
https://www-purepc-pl.translate.goo..._sl=pl&_x_tr_tl=en&_x_tr_hl=pl&_x_tr_pto=wapp
Fun fact is that L4 cache on Broadwell was not really designed/dedicated to CPU. It was there for iGPU. Intel also didn't really advertise this feature. And still it dramatically improved gaming performance.

I don't see any reason why L4 won't improve performance on Nova Lake just like it does on X3D.
How much improvement it will bring and how it will compare to X3D (especially Zen6) of course depends on how good these CPUs will be at gaming nand how good will their extra cache implementation be.

Its really not that bad. With the fixes in place and 200s boost implemented, arrow lake is crazy good value.
I didn't see any tests of any fixes which would bring any improvement.
Arrow Lake despite being worse than competition (including RPL) isn't just that bad at gaming. Especially if you consider actually useful screen resolutions.

Zen 1 and Zen 2 were terrible at gaming compared to Skylake and no one considered these CPUs a joke where it came to gaming with many people claiming they were pretty great overall.
Regressions in performance are always more highlighted.
 
Intel already had CPUs with large cache. Broadwell with 128MB L4 cache in Broadwell and it @4.2GHz often beat latter Skylake CPUs 5Ghz in games.
https://www-purepc-pl.translate.goo..._sl=pl&_x_tr_tl=en&_x_tr_hl=pl&_x_tr_pto=wapp
Fun fact is that L4 cache on Broadwell was not really designed/dedicated to CPU. It was there for iGPU. Intel also didn't really advertise this feature. And still it dramatically improved gaming performance.

I don't see any reason why L4 won't improve performance on Nova Lake just like it does on X3D.
How much improvement it will bring and how it will compare to X3D (especially Zen6) of course depends on how good these CPUs will be at gaming nand how good will their extra cache implementation be.


I didn't see any tests of any fixes which would bring any improvement.
Arrow Lake despite being worse than competition (including RPL) isn't just that bad at gaming. Especially if you consider actually useful screen resolutions.

Zen 1 and Zen 2 were terrible at gaming compared to Skylake and no one considered these CPUs a joke where it came to gaming with many people claiming they were pretty great overall.
Regressions in performance are always more highlighted.
I remember buying the broadwell E 6900K because it was 8 core and had a very big cache at the time. When it was released it was over $1000. It was such a smoooooth gaming experience I loved it.
 
I remember buying the broadwell E 6900K because it was 8 core and had a very big cache at the time. When it was released it was over $1000. It was such a smoooooth gaming experience I loved it.
Broadwell-E didn't have L4 cache though. It was marginally faster than Hashwell-E due to improvements in architecture and clocks afforded by 14nm process node.
Not sure OC but I would assume Broadwell-E didn't have the same 4.2-ish GHz limits as normal Broadwell because it was result of L4 cache not clocking high enough.
No other CPU other than normal consumer Broadwells had this cache.
 
I think the advantage of the 14nm+++++++++++ era was that you knew in a way what you were buying. I mean if reliability was a plus for you then you felt like you could go for it. So if they do eventually come out with Nova Lake, they could do some + models in the future so people can go with what's familiar (or like Bartlett-like models with extra cache and P-core only, even in Ultra 3 and 5 variations). :)
 
KInda what I suspected it's pushed back to 2027 for the Motherboards and the Chips.
could be a videocard typo/copy pasted table, do not see any date mentionned by the source:
https://x.com/jaykihn0/status/2020799854331441293/photo/1

no mention here:
https://overclock3d.net/news/cpu_mainboard/intel-900-series-nova-lake-cpu-chipset-specs-leak/

And yesterday:
https://videocardz.com/newz/intel-r...nd-z970-chipsets-for-nova-lake-s-desktop-cpus
They were still on the 2026 timeline

Would not be a big surprise for an intel delay, but would be really disapointing as they seem to have some nice tech if they can hit yield and release window to shake things up.
 

Top Intel "Nova Lake" Chip Can Pull up to 700 W at Max Load with Power Limits Disabled

by btarunr Today, 05:06 Discuss (10 Comments)
The top Intel Core Ultra Series 4 "Nova Lake-S" desktop processor model, which comes with a total core-count of 52 and implements a big last-level cache (bLLC), could pull an astonishing 700 W in maxed out loads with power limits removed, kopite7kimi reports. The chip has a core-configuration of 16 P-cores, 32 E-cores, and 4 low-power island E-cores. For reference, a current Core Ultra 9 285K with Extreme power profile can pull up to 490 W PL4 in short bursts.

In related news, leaks by Jaykihn shed light on the way clock speeds of the various CPU core types are handled. The base frequency affects the clock speeds of P-cores and E-cores located in the Compute complex, but not those of the low-power island E-cores. "Nova Lake-S" will be Intel's first desktop processors to implement low-power islands, allowing the processor to clock-gate almost the entire compute complex when the PC is idling.”
 
I better buy the 1200 watt Corsaid RMx that is on sale right now...
 
  • Like
Reactions: erek
like this

Top Intel "Nova Lake" Chip Can Pull up to 700 W at Max Load with Power Limits Disabled

by btarunr Today, 05:06 Discuss (10 Comments)
The top Intel Core Ultra Series 4 "Nova Lake-S" desktop processor model, which comes with a total core-count of 52 and implements a big last-level cache (bLLC), could pull an astonishing 700 W in maxed out loads with power limits removed, kopite7kimi reports. The chip has a core-configuration of 16 P-cores, 32 E-cores, and 4 low-power island E-cores. For reference, a current Core Ultra 9 285K with Extreme power profile can pull up to 490 W PL4 in short bursts.

In related news, leaks by Jaykihn shed light on the way clock speeds of the various CPU core types are handled. The base frequency affects the clock speeds of P-cores and E-cores located in the Compute complex, but not those of the low-power island E-cores. "Nova Lake-S" will be Intel's first desktop processors to implement low-power islands, allowing the processor to clock-gate almost the entire compute complex when the PC is idling.”

Hey LukeTBK this bad news? Called it Nova Lake for a solid reason I think. I don't even want the chip if it's going to max out my 850 watt PSU and the 9 other things I have connected to my power strip in my 1960s parents house. I just want to know the base TDP if it's practical enough.
 
Last edited:
  • Like
Reactions: erek
like this
I better buy the 1200 watt Corsaid RMx that is on sale right now...
Hey LukeTBK this bad news? Called it Nova Lake for a solid reason I think. I don't even want the chip if it's going to max out my 850 watt PSU and the 9 other things I have connected to my power strip in my 1960s parents house. I just want to know the base TDP if it's practical enough.

Intel Core Ultra 400 "Nova Lake-S" NPU Churns Out 74 TOPS Throughput

by btarunr Today, 11:32 Discuss (0 Comments)
Intel's upcoming Core Ultra Series 4 "Nova Lake-S" could be the company's first desktop processor series to offer Microsoft Copilot+ native acceleration capability. The NPU of "Nova Lake" offers throughput of 74 TOPS, far exceeding the 40 TOPS required by Copilot+. Intel already met this with its "Lunar Lake" mobile processors, but its current "Arrow Lake-S" and upcoming "Arrow Lake-S Refresh" processors only do 13 TOPS as they implement an older generation NPU carried over from "Meteor Lake."

This is not to say that all desktop processor models of "Nova Lake-S" offer 74 TOPS, some could offer lower throughput, but there's plenty of product segmentation room for Intel to ensure that even its cheapest retail model offers at least 40 TOPS to meet Copilot+ requirements. AMD might end up beating Intel to the title of first socketed desktop processor with a Copilot+ capable NPU. The company could launch its Ryzen AI 400 series "Gorgon Point" Socket AM5 APUs sooner than Intel's rumored late-2026 debut of the Core Ultra Series 4 desktop processors.“
 
Hey LukeTBK this bad news? Called it Nova Lake for a solid reason I think. I don't even want the chip if it's going to max out my 850 watt PSU and the 9 other things I have connected to my power strip in my 1960s parents house. I just want to know the base TDP if it's practical enough.
i think it is between meaningless and small good news.

If i remove all power limit, the more watt the cpu is able to handle the better (why not ? mean good substrate/chip/power delivery I guess, will not mind regular people OC) or at least no idea what it could mean if it is lower or higher. It has 52 cores, 285k with half of them is getting close to 500 watt, so I imagine it is just pretty much the same has last gen ?
 
It's just a better chip for those who need it for that purpose. I think the 14700KS still holds the record for the highest overclock. If the TDP is the same as last Gen I'm fine with that.
 
  • Like
Reactions: erek
like this
1770755079991.png
 
not with PL in place (of course....) it is a dual-cpu enthusiast-HEDT system after all, like shown in the picture, that not something any regular people would need or able to use much, if it is two different giant L3 cache you really need a workload that work well on separate tile to use it so probably not even a better if money is not an object at gaming.

no quad channel, but with ddr-5 speed it support, still a lot of bandwith, w980 workstation board will have ecc support/48 pci lane.
 
Last edited:
There's no way they release a CPU for home use, which legitimately uses 700 watts.

There is virtually no off the shelf consumer product which can cool that much power, from a CPU with an IHS. You would have to de-lid and do custom water cooling with a lot of water volume, to keep that thing reasonably cool.

And also, I would question the durability of typical motherboard sockets and VRMs, to handle that much power sustained.
 
Yeah I mean if it was possible for it to juice 700 Watts there's no way I would allow it I probably cut off half or if not all of the ecores and tune the 16p cores properly and maybe just allow eight e cores or 16 ecores with it 32 ecores is over the top
 
This is one of those clickbait headlines.

700W is supposedly the "PL4" hard upper limit for the dual compute tile 52-core model with power limits "disabled". The 285k's hard limit is ~490W. If you think of it as 2x350W or in terms of watts per core it's quite a bit lower than Arrow Lake. Almost no one will run these things that way. 700W would be way into diminishing returns, and particularly useless for gaming if Arrow Lake is any indication. 285k rarely hits 200W in a game. Just because max is north of 400W doesn't mean you have any use for it.

I'm bet we'll see the usual 125W "TDP"/official PL1 and 250W PL2 on the Core Ultra 7s. u9s with 52 cores will presumably go higher at least for PL2 & probably PL1, but I bet Intel will prod or require board vendors to default the limits to something a decent AIO can handle. I expect gaming boards will just run them at PL2 all the time by default like they do with Arrow Lake.

The big thing I'm really hoping to see this generation now that Intel is going nuts with the core count is a Core Ultra 7 that clocks at least as high as the 9s. Call it a KS or whatever. 52 cores on 2 compute tiles is going to be a total waste for a gaming rig. Can we not have to buy that to get max clock speed?
 
Screenshot_20260211_015049_Chrome.jpg


Kopite could be spreading fake news. I'm not even sure he works for Intel.
 
Last edited:

Intel "Nova Lake" Compute Tile Die-sizes Surface

by btarunr Today, 06:51 Discuss (0 Comments)
Intel is preparing to launch its next generation Core Ultra Series 4 "Nova Lake" processors in the second half of 2026. We have some hint on what the sizes of the Compute tiles of these chips could measure. "Nova Lake" remains a tile-based disaggregated chip, just like the current "Arrow Lake," with the most advanced foundry node being allocated only to the components that can benefit the most from it, the CPU cores. The CPU complex of the processor is located in the Compute tile, while low-power island E-cores are located on the SoC tile that's built on a slightly older foundry node.

Intel is building the Compute tile of "Nova Lake" on the TSMC N2 (2 nm nanosheet) foundry node. There will be different types of Compute tiles based on the CPU core count and cache sizes. The two most popular ones for the desktop platform will be the mainstream tile with 8P+16E core configuration and a standard L3 cache size shared among the eight P-cores and four E-core clusters. The second, more premium Compute tile type will be 8P+16E with bLLC, or big last-level cache. bLLC is an enlarged L3 cache that's estimated to be 3-4 times the size of the regular L3 cache. This is Intel's answer to AMD's 3D V-Cache. Estimates put the regular Compute tile to measure around 110 mm², while the premium variant with bLLC is estimated to measure over 150 mm². This 36% increase in die-area is entirely from the enlarged last-level cache.“
 
if 150mm was considered too big for Intel 18 yield...

300mm of tsmc N2 + graphic tile + SOC/IO, you end up with a massive package, this will cost a lot, will be interesting to see how their chiplet talking to each other work well
 
Intel declared a dual sourcing strategy for Nova Lake in November 2024... https://www.tomshardware.com/pc-com...rom-intel-nova-lake-is-the-likely-application That article is from April 2025.

Makes sense to me. Back then it would have been a hedge against 18A problems. Even if 18A works great by the time Nova Lake goes into production it could still be useful. Intel is short on capacity for server chips & 18A is also planned for next gen Xeons, so that could be part of it. Dual sourcing could also be useful for dealing with tariffs. Intel's 18A capacity is all in the US. TSMC N2 can be fabbed in Taiwan, packaged outside the US, and could thus be useful for dodging trade wars. It would be particularly useful selling to China. China treats products from Taiwan as domestic, and imposes duties on chips based on where they were fabbed.
 
TSMC N2 can be fabbed in Taiwan, packaged outside the US,
New Mexico I think for advanced Intel packaging, they will malaysia capacity in the future has well, so maybe there will some of them there, but for the 3d part probably new mexico.

The Xeon 18A also has like panther lake small compute tile, which is they have yield issue (or fear them would make sense), only using 18A on small chiplet, tsmc 2N once they get bit., small low-end budget nova lake will also go on intel 18A I think, they have very small chiplet
 
Will there be an ITX variant for this CPU? Or are these CPU sockets getting too big for ITX I wonder if not now in the future
 
Will there be an ITX variant for this CPU? Or are these CPU sockets getting too big for ITX I wonder if not now in the future
that 52 cores nova lake will be on the regular/standard desktop socket (LGA 1954) and it is a regular size package (same as 12th gen to now so it should be doable, (45mmx37.5mm), maybe you will not be able to push it fully over the 300 watt it will be able to use
 

Intel Core Ultra 400 "Nova Lake-S" NPU Churns Out 74 TOPS Throughput

by btarunr Today, 11:32 Discuss (0 Comments)
Intel's upcoming Core Ultra Series 4 "Nova Lake-S" could be the company's first desktop processor series to offer Microsoft Copilot+ native acceleration capability. The NPU of "Nova Lake" offers throughput of 74 TOPS, far exceeding the 40 TOPS required by Copilot+. Intel already met this with its "Lunar Lake" mobile processors, but its current "Arrow Lake-S" and upcoming "Arrow Lake-S Refresh" processors only do 13 TOPS as they implement an older generation NPU carried over from "Meteor Lake."

This is not to say that all desktop processor models of "Nova Lake-S" offer 74 TOPS, some could offer lower throughput, but there's plenty of product segmentation room for Intel to ensure that even its cheapest retail model offers at least 40 TOPS to meet Copilot+ requirements. AMD might end up beating Intel to the title of first socketed desktop processor with a Copilot+ capable NPU. The company could launch its Ryzen AI 400 series "Gorgon Point" Socket AM5 APUs sooner than Intel's rumored late-2026 debut of the Core Ultra Series 4 desktop processors.“
72 TOPS would be the last reason I would want this CPU as my current setup does +1000 TOPS but 16 P-cores (no HT) and 32 E-cores would be a reason combined with a RTX 6090 for a worthy upgrade.
 
72 TOPS would be the last reason I would want this CPU as my current setup does +1000 TOPS but 16 P-cores (no HT) and 32 E-cores would be a reason combined with a RTX 6090 for a worthy upgrade.
If it was only 8 pcores in 32 ecores I'd be extremely disappointed but since they're 16 p cores now we're talking 👍I will most likely upgrade to the 16 pcore version
 
I don't know if this applies to Nova Lake, but I'm a bit confused. Does this mean P-core only models in the future?
more a successor to the e-core than the p-core architecture only, all the cores will share the same ISA, which is different than no what would be called performance or big core (and island low pwoer cores could continue) vs smaller one

More going AMD ways of doing things with the zen5c vs zen5 cores could be what they go for.
 
Rumored Nova Lake bLLC SKUs:
Core Ultra X9 400X (HEDT)

52 cores (16+32+4 LPE), up to 288MB bLLC
44 cores (16+24+4 LPE), up to 288MB bLLC
42 cores (14+24+4 LPE), up to 288MB bLLC


Core Ultra 7 400K (unlocked)

28 cores (8+16+4 LPE), up to 144MB bLLC
24 cores (8+12+4 LPE), up to 144MB bLLC
Core...

Source: VideoCardz.com
https://videocardz.com/newz/intel-n...e-sku-reportedly-upgraded-from-42-to-44-cores


This thing is being fabricated in Arizona should be interesting hope it lauches this year or early next.
 
Last edited:
Very curious if the hedt chips are faster than the 400k i7 chips in gaming. Also no i9s for non hedt? And of course what will the price be for the 16 pcore 44 core chip looks like the sweet spot perhaps.
 
Very curious if the hedt chips are faster than the 400k i7 chips in gaming.
I would guess not because the core ultra 7 400k still get the full 144MB of cache according ot this, the x9 are just the double die and a game will want/benefit to stay all on the same die. Intel having historically larger L2 and tax less the L3 will have to see if those giant cache boost gaming performance the same way it did for AMD, it could be quite the generational uptick if it do.
 
Last edited:
I would guess not because the core ultra 7 400k still get the full 144MB of cache according ot this, the x9 are just the double die and a game will want (benefit) to stay all on the same die.
Add to it the extra heat from all those extra cores, and it probably can't clock as high - and frequency does still help with games, especially if the frequency disparity is large.
 
Rumored Nova Lake bLLC SKUs:
Core Ultra X9 400X (HEDT)

52 cores (16+32+4 LPE), up to 288MB bLLC
44 cores (16+24+4 LPE), up to 288MB bLLC
42 cores (14+24+4 LPE), up to 288MB bLLC


Core Ultra 7 400K (unlocked)

28 cores (8+16+4 LPE), up to 144MB bLLC
24 cores (8+12+4 LPE), up to 144MB bLLC
Core...

Source: VideoCardz.com
https://videocardz.com/newz/intel-n...e-sku-reportedly-upgraded-from-42-to-44-cores


This thing is being fabricated in Arizona should be interesting hope it lauches this year or early next.
Core Ultra X9 485X(?) (52 cores (16+32+4 LPE), up to 288MB bLLC) + an RTX 6090 will be my next upgrade.

Will basically be a doubling of my current setup CPU wise: from 8(HT)+16 -> 16+32+4 LPE and then 2 GPU generation liftup. (RTX 4090 -> RXT 6090).
Only question is if I am going from 64 GB DRR5 6600 - 128 GB DDR 58800.
 
The high core count is also a problem with the high RAM prices. If you use the cores for applications that use a good chunk of memory per scheduled entity (such as compilaton) you need sufficient RAM to keep them all fed. A smaller number of faster cores saves money on RAM for the same total throughput.
 
The high core count is also a problem with the high RAM prices. If you use the cores for applications that use a good chunk of memory per scheduled entity (such as compilaton) you need sufficient RAM to keep them all fed. A smaller number of faster cores saves money on RAM for the same total throughput.
Price is the least of my factors I build a PC to last ~4 years and then I build a new PC so when I divide the cost up over 4 years, the price is kinda low as it makes a $6200 PC into $1550 per year.
 
Back
Top