Intel Core Ultra 200 "Arrow Lake-S" Desktop Processor Core Configurations Surface

[JustBenching]

Doing some really rough napkin math, strictly related to cb23, it’s looking really bad for MT performance on the i9 (~36,000-37,000). With losing HT, not even the 50% IPC improvement to the ecores is going to make enough difference. I hope there’s some extra sauce due to negated mitigations from HT removal.

With no power or thermal throttling 13900k scores 41200. With HT off it does 36k. A 50% boost to the ecores should net around 10k points.

 

[rv8000]

With no power or thermal throttling 13900k scores 41200. With HT off it does 36k. A 50% boost to the ecores should net around 10k points.

A stock 14900K scores 38476 (as per TPU review). 4 ecores account for roughly 3900 points (*4 that would be 15600).

38476 - 15,600 = 22876 (removal of ecore score).

22876 - 5200 (your number for disabling hyper threading, I assumed larger deficit previously) = 17676 (base pcore)

(17676 *1.10) + (15600 * 1.5) = 41,076

Without knowing final clock frequencies (which are as of intels mobile release guaranteed to be lower) and what gains from potential mitigations due to hyper threading removal, intel will barely surpass 14th gen in MT. I hope they focus more on improving efficiency and stock TDP, as both 13th and 14th gen are horrendous. I wouldn’t be bothered by a marginal shift in MT, but compared to the 9000 series, potentially not releasing until December is likely to make this new line of processors look poor vs the competition.

 

[JustBenching]

A stock 14900K scores 38476 (as per TPU review). 4 ecores account for roughly 3900 points (*4 that would be 15600).

38476 - 15,600 = 22876 (removal of ecore score).

22876 - 5200 (your number for disabling hyper threading, I assumed larger deficit previously) = 17676 (base pcore)

(17676 *1.10) + (15600 * 1.5) = 41,076

Without knowing final clock frequencies (which are as of intels mobile release guaranteed to be lower) and what gains from potential mitigations due to hyper threading removal, intel will barely surpass 14th gen in MT. I hope they focus more on improving efficiency and stock TDP, as both 13th and 14th gen are horrendous. I wouldn’t be bothered by a marginal shift in MT, but compared to the 9000 series, potentially not releasing until December is likely to make this new line of processors look poor vs the competition.

TPU doesn't use realtime priority so you can't cross compare our scores. I can assure you a stock 13900k with realtime priority (to remove any background tasks noise) scores around 41k to 42k. 16 ecores definitely don't score only 15k. 8pcores @ 3.7 ghz (alderlake) score around 9k+. So the 16 on a 13900k running at 4.3 should be at around 20.

Efficiency is already great, stock TDP sure, but that's configurable so who cares.

 

[rv8000]

TPU doesn't use realtime priority so you can't cross compare our scores. I can assure you a stock 13900k with realtime priority (to remove any background tasks noise) scores around 41k to 42k. 16 ecores definitely don't score only 15k. 8pcores @ 3.7 ghz (alderlake) score around 9k+. So the 16 on a 13900k running at 4.3 should be at around 20.

Efficiency is already great, stock TDP sure, but that's configurable so who cares.

I shouldn’t expect any less of you.

TPU scores are what I find trustworthy and also as a baseline. You can cherry pick and move the goal posts as much as you want, the fact remains what you attempt to pass off as “default” or baseline scores for any Intel processor across the forums is always modified through bios alterations - this is not stock.

Either way, you can choose to believe whatever you like, it’s an estimation based on what we know so far: IPC improvement, clock speed regression, lack of hyper threading, and a best case scenario that the e-cores are in fact 50% faster.

We don’t need another thread with you spreading delusions about their efficiency, or lack there of.

 

[JustBenching]

I shouldn’t expect any less of you.

TPU scores are what I find trustworthy and also as a baseline. You can cherry pick and move the goal posts as much as you want, the fact remains what you attempt to pass off as “default” or baseline scores for any Intel processor across the forums is always modified through bios alterations - this is not stock.

Either way, you can choose to believe whatever you like, it’s an estimation based on what we know so far: IPC improvement, clock speed regression, lack of hyper threading, and a best case scenario that the e-cores are in fact 50% faster.

We don’t need another thread with you spreading delusions about their efficiency, or lack there of.

What the hell are you talking about?

You are using my scores with HT off and cross compare it with TPUs when we are testing with different methodologies. That's all I pointed out. You can't use my HT off scores and compare them with TPUs..

If you don't think RPL are efficient then you are the one with the delusions bud. 14900k is the 2nd most efficient desktop chip in MT efficiency. If that's not efficient enough for you then what the hell is every other chip?

You said that Intel should decrease TDP and increase efficiency. I'm pointing out that if they decrease their TDP, their efficiency is already superb on 14th gen. Here is the data that demonstrates it

Spoiler

 

[ShrekOC]

As expected, it also comes with a new type of socket.
Never understand why they change the socket type very often.

It's of utmost importance that the pins get slightly rearranged every 2-3 years