Some Intel Nova Lake CPUs Rumored to Challenge AMD's 3D V-Cache in Desktop Gaming

[Hecate91]

I can see this thread getting locked again. Intel is in the rear view in DIY already. As teenagers who have been with AMD since 2017 will lead to an increase in AMD market share. People who are ragging on AMD can thank them for 6 and 8 core CPUs. If Intel were in the lead you would still be getting 4 core CPUs in the Consumer market. Then there is the truth. By the time Nova Lake launches AMD will have something to respond. I am amazed at how much faster the 9900X3D is vs the 7900X3D and that was already a fast chip. One thing it proves is Intel have to respond to X3D as it was them that started the narrative in DIY when the 1700X launched with 8 cores all those years ago. The best though I won't need to change my MB.

Intel being in the rear view is well deserved for them after years of stagnation and market manipulation through oem's. It isn't surprising the Intel fans still want to bash on X3D even though it has been very successful, so successful Intel wants to copy the tech.
If Intel were in the lead we would probably have 6 cores at most and they'd still be on 10nm, with ridiculous segmentation on high end desktop.
By the time Intel responds with Nova lake, it will require a new motherboard, while anyone with a b650 or x670 motherboard can have a drop in upgrade from Zen 4.

 

[A Computer Guy]

Sure, here you go. 28k at 95w, the 9700x barely scores 22

Intel Core Ultra 7 265K cooling testing: How much does it take to keep Arrow Lake cool in MSI's MPG Velox 100R PC Case?

Intel’s Core Ultra 7 265K is easier to cool than previous-generation CPUs like the Core i7-14700K.

www.tomshardware.com

What you posted shows about a 22% reduction in performance for the 265k at 95w per the article that was about cooling using CB23 as a metric and not discussing 9700x at all.

For the sake of argument if I apply that % to TPU's scores CB2024 and make some assumptions then for 265k that is about 1630pts (from 2090pts) compared to 9700x's 1208pts.

If I apply TPU's points per watt efficiency formula I get
1630pts/95w = 17.2 points per watt
1208pts/80w = 15.1 points per watt (TPU said 15.7)

Before you celebrate your victory let me state my errata
- My assumption in CB scaling between CB23 and CB2024 may be completely wrong.
- How CB stresses the chips power draw between CB23 and CB2024 may be different.
- Comparing power measurements between TOM's and TPU's reviews may not be compatible.
- TPU ppw I calculated suggests CB2024 pulled 76w-77w not 80w which I got from a different part of the review.
- 3pt spread might be within a margin of error.
- Maybe a messed up a calculation.

Final Thoughts
If you want to run a 256k daily at 22% reduced performance you might as well get a 9700x and enjoy one last socket upgrade when AM5 goes EOL.

 

[Squared]

By the time Nova Lake launches AMD will have something to respond.

Will AMD have a response at the same time? Some rumors say N2X for Zen 6, and N2P for Nova Lake, which should mean Nova Lake's node is available first. But surely they'll compete with one another eventually.

Intel had to respond to the 1700x, when their existing 4 core cpus were better in almost eveything due to the superior IPC?

I bought an 1800X because it had similar efficiency, worse single-core performance (including games), but far better multi-core performance than Intel. Today the tables have turned. Today I'd consider something like the 265K, as it's very cheap ($40 cheaper than a 9700X), matches AMD in efficiency and single-core performance, roughly matches AMD in gaming performance (within 4% of 9700X, but loses by a huge margin against the much more expensive 9800X3D), and crushes AMD in multithreaded performance (competes with the 9900X for less than the 9700X). The only thing that would give me pause is that last time I accepted like 15% less gaming and low-thread performance, but Arrow Lake loses by around 30% in gaming.

Thankfully I don't need a new CPU today. Passing up X3D would be hard, but accepting only 8 cores per L3 cache block would also be hard. My 1800X kept up for a long time because games became more threaded. What happens to the X3D chips when future games spill over into more cores? Arrow Lake has continuous L3 so it looks more future-proof but it doesn't have AVX-512. (I did replace my 1800X with a 5700X3D. That was cool but I don't think I gained enough to consider doing it again; next time I'm waiting longer between upgrades than even the life of AM4.) Nova Lake with bLLC spounds like it'd be an easier choice.

 

[JustBenching]

What you posted shows about a 22% reduction in performance for the 265k at 95w per the article that was about cooling using CB23 as a metric and not discussing 9700x at all.

For the sake of argument if I apply that % to TPU's scores CB2024 and make some assumptions then for 265k that is about 1630pts (from 2090pts) compared to 9700x's 1208pts.

If I apply TPU's points per watt efficiency formula I get
1630pts/95w = 17.2 points per watt
1208pts/80w = 15.1 points per watt (TPU said 15.7)

Before you celebrate your victory let me state my errata
- My assumption in CB scaling between CB23 and CB2024 may be completely wrong.
- How CB stresses the chips power draw between CB23 and CB2024 may be different.
- Comparing power measurements between TOM's and TPU's reviews may not be compatible.
- TPU ppw I calculated suggests CB2024 pulled 76w-77w not 80w which I got from a different part of the review.
- 3pt spread might be within a margin of error.
- Maybe a messed up a calculation.

Final Thoughts
If you want to run a 256k daily at 22% reduced performance you might as well get a 9700x and enjoy one last socket upgrade when AM5 goes EOL.

Thats goal post moving. The question was whether or not it was more efficient. Clearly it is.

Also the gap is much larger than what you are calculating cause you are still not doing an iso comparison. The question is in order for the 9700x to hit that 1.6k score to match the 265k, how much power would it need? 200 watts? More than that?

Also why in the world would you buy a 9700x in this case? Thats freaking absurd man, even in yoir example the 265k is 50% faster while also being more efficient, but youd go with amd

 

[Tyler-98-W68]

Intel fans want to bash on x3d, I remember when intel starting loading extra cache on cpus.

I guess some people like to live like it's 20yrs ago and game at 1080p I prefer to game at 4k and after owning 2 x3d chips im still trying to figure out what's so special about them at 4k...

 

[A Computer Guy]

Thats goal post moving. The question was whether or not it was more efficient. Clearly it is.

I didn't move any goal posts. You provided me some data and I estimated that reducing the performance of the 265k by 22% you match the better efficiency of the 9700x per TPU's points per watt efficiency formula. Take it for what it is along with a grain of salt. If it's not they way you want to prove your point then you should put up the effort or retire from the conversation.

Also the gap is much larger than what you are calculating cause you are still not doing an iso comparison. The question is in order for the 9700x to hit that 1.6k score to match the 265k, how much power would it need? 200 watts? More than that?

Your being absurd.

Also why in the world would you buy a 9700x in this case? Thats freaking absurd man, even in yoir example the 265k is 50% faster while also being more efficient, but youd go with amd

Outside of specific constraints/goals (in my opinion) generally it doesn't make sense to give up 22% of 265K's performance to match 9700x's efficiency at the same price point. I see on Amazon they are going for about $300. Wither or not 265K or 9700x is a good fit for you is something you have to decide on your own based on your workload. At $300 it's a very competitive offer.

And to stray this back on topic if it had v-cache on the pcores it could be an awesome gaming product for games that can use that.

 

[JustBenching]

Outside of specific constraints/goals (in my opinion) generally it doesn't make sense to give up 22% of 265K's performance to match 9700x's efficiency at the same price point.

I agree, it doesnt make sense to reduce the 265ks performance by 22% (even though it will still be 50% faster than the 9700x) and that's because efficiency isnt that important when you are sacrificing performance to get there. Which bring us back to my original point, amds chips arent efficient, they are just slow. They are doing exactly what you described, they are murdering performance to keep the power draw low.

An actual efficient chip is one that can remain stupid fast while keeping the power low, like the 265k does with the numbers i gave you. The 9700x just cannot do that cause its not efficient.

 

[A Computer Guy]

I agree, it doesnt make sense to reduce the 265ks performance by 22% (even though it will still be 50% faster than the 9700x) and that's because efficiency isnt that important when you are sacrificing performance to get there. Which bring us back to my original point, amds chips arent efficient, they are just slow.

For the sake of argument the only way to convince me of that is to see real tested comparison with only the pcores so your comparing performance and power using 8 cores vs. 8 cores not 20 cores with 8 cores. Your point stands on nothing proven yet other than a modern Intel 20 core cpu continues to beat a modern AMD 8 core cpu in CB at about the same power efficiency. Anybody could have guessed that to be the case because 20 > 8.

An actual efficient chip is one that can remain stupid fast while keeping the power low, like the 265k does with the numbers i gave you. The 9700x just cannot do that cause its not efficient.

Those numbers addressed full multi-core workload only between a 20 vs. 8 core CPU. Your reasoning is still unproven.

 

[Panther_Seraphin]

Delusional take. Until 2022 with thr first x3d chips amd was behind not just in gaming performance but also in gaming power draw. Amd fans....

The 10 and 11900k doesnt exist in your Universe huh.......

You just confimed my point as well as you have to be super specific to make intel win vs AMD. Flip the question. How did that Intel CPU do when doing something in productivity?

And to stray this back on topic if it had v-cache on the pcores it could be an awesome gaming product for games that can use that.

I wonder how much Intel will benefit from X3D like tech due to their better memory controller?

 

[JustBenching]

For the sake of argument the only way to convince me of that is to see real tested comparison with only the pcores so your comparing performance and power using 8 cores vs. 8 cores not 20 cores with 8 cores. Your point stands on nothing proven yet other than a modern Intel 20 core cpu continues to beat a modern AMD 8 core cpu in CB at about the same power efficiency. Anybody could have guessed that to be the case because 20 > 8.

Those numbers addressed full multi-core workload only between a 20 vs. 8 core CPU. Your reasoning is still unproven.

The comparisons are done on pricepoints, not on core counts. Core counts are meaningless. Your basically saying that amd is slower and less efficient cause it has less cores. Oh well, nobody is stopping them from offering more or dropping the prices so their 8 core part doesnt cost as much as intels 20 core part. Until they do that intel will remain most efficient at the mid and low range.

The 10 and 11900k doesnt exist in your Universe huh.......

You just confimed my point as well as you have to be super specific to make intel win vs AMD. Flip the question. How did that Intel CPU do when doing something in productivity?

The 10900k wasnt pulling more power than the 5950x did in games. I have no clue about 11,those chips were garbage.

 

[blinnbanir]

The 10900k wasnt pulling more power than the 5950x did in games. I have no clue about 11,those chips were garbage.

Total BS

 

[A Computer Guy]

The comparisons are done on pricepoints, not on core counts.

LOL you just moved the goal post.

Core counts are meaningless.

Your basically saying that amd is slower and less efficient cause it has less cores.

I'm saying you haven't made your point because your comparison is garbage because core count matters in the performance/power comparison.

Oh well, nobody is stopping them from offering more or dropping the prices so their 8 core part doesnt cost as much as intels 20 core part. Until they do that intel will remain most efficient at the mid and low range.

Have a great day.

 

[Panther_Seraphin]

The 10900k wasnt pulling more power than the 5950x did in games. I have no clue about 11,those chips were garbage.

5950x less power than 10/11/12th gen in gaming on average

 

[JustBenching]

'm saying you haven't made your point because your comparison is garbage because core count matters in the performance/power comparison.

Yes, corecounts do matter, that's why intel is more r efficient, cause they offer more cores. Anyways, im out,this is madness.

 

[Rover4444]

Wow! I sure am glad I'm not caught in another shitty argument that spans three pages! Thank you, ignore button!

Just buy the Microcenter 265K bundle, guys. CUDIMM support, Z890 boards, and $200 less than the X670E 9800X3D combo. No need to argue

 

[Outback Bronze]

Sorry guys take a break. This thread has flown off the rails.