Intel Ultra 5 250K Plus CPU Review & Benchmarks: Gaming, Production, & Power Consumption

0:01 All right, Intel has a CPU.

0:03 It's a $200 CPU.

0:05 It's called the Ultra 5 250K Plus, the 250KP as we're calling it.

0:09 And there's some software that you need to use

0:11 with it called the Intel Platform Performance Package,

0:15 or as I've been calling it, Intel's PP package.

0:18 Now, the Intel PP package isn't too big.

0:21 It's only about 100 megabytes, but it does seem to get the job done for Intel.

0:25 Uh, when we were testing Intel's PV package,

0:28 I'm not sure why my team kept laughing like every single time I was asking them,

0:32 did you use the PV package or you forgot to use the PV package?

0:36 And I every they just kept laughing at me.

0:38 I don't know.

0:39 I'm not I don't really get it.

0:40 It just felt like we're all just going

0:42 to laugh at Steve today for apparently no reason

0:44 whatsoever when all I was trying to do is

0:47 make sure that Intel's PP package is getting use.

0:50 Thanks, Steve.

0:51 Anyway, the company is launching a $300 270K Plus

0:54 CPU as well that we'll review in our next video.

0:57 These two are the Aerol Lakeake refresh CPUs

1:00 that use the existing LGA1 1851 socket motherboards.

1:03 I'll give you the TLDDR to save everyone a little bit of time.

1:05 The short version up front is that Intel's new 250K Plus

1:08 CPU finally brings Intel back to parody with its 14th gen,

1:11 undoing a lot of the regression and gaming

1:13 that we saw with the 200 series at launch.

1:15 In some situations, it even best the 285K by a couple FPS,

1:19 although enabling 200s boost on the 285K can sometimes equalize them.

1:24 Again, most of the time,

1:24 we see the 250K Plus around the level of the 14600K and the 285K.

1:29 It trails the 9800 XD predictably and significantly,

1:31 but it also costs a lot less, so it isn't a direct comparison.

1:34 We're seeing large gains in gaming over

1:35 the 9600X and actually production workloads as well, like 24% in this case,

1:40 and significantly better performance than Intel's

1:42 own preceding 245K in basically every situation.

1:45 That's largely due to going with more cores and cache for the 250K

1:48 plus rather than the 245K as opposed to a pure clock bump.

1:51 In non- gaming workloads,

1:52 Intel maintains a good overall position against AMD's price

1:55 equivalent options like the 9600X or the 250K Plus.

1:59 For the most part in the test that we run,

2:00 Intel 250 KP is just straight up better in a lot of the non-gaming

2:04 tests and actually in a lot of the gaming tests versus the 9600 X.

2:07 That said, AMD's 5800 X3D and 5700 X3D 5600 X3D still

2:11 do really well in gaming and often do beat the 250 KP.

2:15 The biggest challenge for Intel remains the fact

2:17 that Intel has a poor history of motherboard longevity.

2:20 This is probably the last in the line other

2:22 than maybe other AeroLake refreshes for this motherboard socket.

2:25 We're keeping this one pretty simple today.

2:27 Still a lot of charts.

2:28 There's a lot of data, but relatively simple as far as we're concerned.

2:31 We're not going to get into the efficiency

2:33 as in performance per watt that we've done in the past.

2:35 And to be real open with you, I am tired.

2:40 Like I have not left the office for about a week

2:43 and I would like to go sleep at some point.

2:45 So, uh, the couch is getting its I mean, it's not great for my back.

2:49 So, we're going to keep this focused on gaming production frequency.

2:53 Got a little bit of power in there just to address it.

2:55 Uh, and then maybe we'll expand the scope on this in the future.

2:58 If there's some interest in these parts,

2:59 we can always add some additional tests.

3:01 But for the most part, I think we've got a pretty good scope of coverage here.

3:04 Uh, we'll start with the 250KP today.

3:06 We'll do the 270KP in a separate review video coming

3:09 up basically immediately after this one as our next video.

3:12 Um, and the the reason this has been such

3:15 a haul is not because of the CPUs necessarily,

3:17 but because in the last week we had the DLSS5 news.

3:20 We had Noct was case review that we ran where uh

3:24 that is the most in-depth case benchmark I think we've ever done.

3:27 We had the fan testing data in there as well.

3:29 We ran Discord alternatives.

3:30 We did the Crimson Desert benchmarking.

3:32 We debuted our animation error or simulation time

3:35 error charts for actual GPU comparison in that testing.

3:38 So, it has been a very busy week,

3:40 but we're going to close it out with the 2D KP review.

3:43 Let's get started.

3:44 Before that, this video is brought to you by the Mont HSO1 and HSO2 cases.

3:48 The HSO1 and HSO2 cases differ by way of a mesh front or a curved glass front.

3:53 We previously reviewed these cases

3:55 in our intensive benchmarking process and found them

3:57 to be overall high build quality

3:58 with thermals comparable between the two designs.

4:01 The rear chamber of the case uniquely brings

4:03 the power supply forward and perforates the rear

4:05 wall of the case allowing for more

4:07 air movement internally while providing a different look.

4:09 The rear of the case uses a fan mounting mechanism for two

4:12 separate fans that can be coupled together with a plate improving convenience.

4:16 The HSO1 and HSO2 also invert with minimal effort,

4:20 allowing the computer to present to the opposite

4:22 side without needing to reconfigure the entire build.

4:25 Learn more at the links in the description below.

4:27 The new CPU has launched in a few days, which is good.

4:28 That means they're giving people time

4:30 to watch reviews before they make a decision.

4:32 So, that's nice.

4:32 Uh, behind the scenes, Intel was extremely disorganized,

4:35 at least in handling our side of things, for the reviews for this one.

4:38 So, I I haven't really seen Intel this scatter.

4:41 They were late with getting us information.

4:43 They're late getting software out.

4:45 the CPUs didn't really arrive with that much time to work

4:47 on them either for us and I I just was not impressed.

4:51 Now that said, that's not part of the product.

4:52 It's not part of the consumer experience.

4:54 So, we won't factor that into the conclusions or anything,

4:57 but I do think it's maybe an indicator

4:58 of where Intel's at where this is not like Intel.

5:00 I don't know if it's because of the layoffs or what, but anyway,

5:03 we'll start with a refresher on the CPU side of the market.

5:06 We already know RAM is rag.

5:07 So, focusing on the CPUs, the 250 KP is a $200 part.

5:10 Currently, the 245K is also $200.

5:12 The 14600 K is largely gone other than third-party ripoffs

5:15 or used listings which are fine but variable in price.

5:18 And these 9600X is currently about $190,

5:20 making it a head-to-head competitor by price.

5:22 The 7700 X is around $250 with the 7800 X3D at around 350 to $380.

5:28 Now, the 7800 X3D is definitely worth considering at that price,

5:31 but maybe more as an alternative to the more

5:33 similarly priced 270KP that we'll review next.

5:35 It's in a different price class for the 250 KP.

5:38 Just for reference, the 285K is currently $530 after a discount

5:41 code with the 9800 X3D at 450 after a similar discount code.

5:46 Intel's new 250K Plus CPU is in the former i5 lineup and should be $190 to $200.

5:51 They say the 250K Plus is running a 6P core 12E core

5:54 configuration for an 18 core 18thread spec at 5.3 GHz max boost advertised.

5:59 TDP is claimed to be 159 watts for the 250K Plus.

6:02 Although TDP and actual power don't match exactly,

6:04 Intel is also now claiming native support for up to DDR5 7200 on the 250 KP,

6:10 up from 6400 previously.

6:12 So great, higherend memory is supported.

6:17 Now you can max out your $200 CPU with your $2,400

6:21 RAM in your motherboard that you'll only get to use once.

6:25 Notably, the extra cores also print with them

6:27 an L2 cache increased to 30 megabytes.

6:30 For comparison, Intel's 245K runs a sixp core AE core

6:33 configuration for a 14 core 14thread spec at 5.2 GHz.

6:37 That's 100 MHz reduced speed for maximum advertised boost

6:40 and that's four fewer ecores than the new CPU.

6:42 This is a reconfiguration of the spec itself.

6:44 So, the 250K Plus is not an exact replacement for the 245K.

6:48 The core count change means it's not only a frequency bump.

6:50 The 245K is on 26 megabytes of L2 also.

6:53 So, while the 250 KP is no X3D,

6:56 it should still have an outsized impact on performance

6:58 against the older 245K with its larger cache.

7:00 For reference, the Intel 285K is in the former i9 lineup,

7:04 now called Ultra 9200S, because that makes more sense somehow,

7:07 and it remains the highest core count spec of this group.

7:10 The 285K has eight pores, 16 EC cores, and 24 total cores and threads.

7:15 Max advertised boost is 5.7 GHz,

7:17 making it the highest of the group with ecore still at 4.6.

7:21 Keep in mind the actualized boost will be

7:23 lower as more threads are engaged with work.

7:25 All right, these CPUs require some additional setup.

7:27 This is important for you to know as well because if you do actually buy one,

7:29 you'll need to do this stuff.

7:30 Intel setup documentation reminded us of AMD's old 30-page core

7:34 parking setup guide from its first generation of Park CPUs.

7:37 And Intel's document reads, quote, exclamation point exclamation point.

7:41 Stop.

7:42 Important information exclamation point exclamation point.

7:46 End quote.

7:47 That's how you know it's uh well I I think they're trying to say it's important.

7:51 They then go on to say that the Intel platform performance

7:54 package or the PP package for short or the PPP for even

7:59 shorter must be installed in order to correctly activate quote scheduling

8:03 core parking idle power performance results and feature availability end quote.

8:09 Intel warns in both bold and underline that a Windows

8:12 system won't work right without this jokes on them.

8:15 Windows never works right.

8:17 The setup guide explains that reviewers received

8:19 an impotent version of Intel's PP package,

8:22 noting that it requires a second .exe.

8:24 In other words, after installing the CPU,

8:26 the management engine, and the chipset drivers,

8:28 we also had to install a separate .exe,

8:30 and you will too, for the platform performance package software.

8:34 This is not automatically pulled by Windows, at least not right now.

8:37 In our case, we also had to install a second

8:40 separate .exe for the driver tuning patch hot fix.

8:43 Intel claims that the public version will only require the first .exe,

8:47 not the hot fix, that they had to push out before the product launched,

8:52 but the fact that Intel made an active decision

8:54 to ship bugs to reviewers is problematic for different reasons.

8:57 Now, we installed both the PB package and the hot fix for it.

9:01 So, we followed the setup guide.

9:02 We did, you know, as you're supposed to do to get the CPU to work properly.

9:05 Um, still though, we don't review the future.

9:08 We don't review promises from companies.

9:10 We review what they send us.

9:11 And so although yes, we did set it up the way Intel says you're supposed to, uh,

9:15 we also have to assume that there's a chance that there's

9:19 some bug issues or that the hot fix doesn't make it

9:22 in or whatever into the consumer products in a couple days because

9:24 I have no way of verifying that from where I am right now,

9:27 which is before launch.

9:28 I just know what they sent me.

9:29 And uh, so it's just disorganized.

9:31 Now, it does at least work if you do everything right,

9:34 but end user reviewer, whatever.

9:37 A company shouldn't be sending out a product.

9:38 It's like you have to install this thing

9:40 and then you have to immediately install this fix

9:41 for this broken thing that you had to install

9:43 to get the thing to work the first time.

9:44 Intel also talked a lot of game about

9:46 its binary optimization tool which it says quote opens

9:49 a parallel to hardware path for Intel to inspect

9:51 and optimize CPU circuit utilization to reduce cache misses,

9:54 branch mispredicts, microarchitectural hotspots,

9:57 and other forms of artificial latency in the compute pipeline.

9:59 End quote.

10:00 Unfortunately, in a throwback to Intel's busted APO launch,

10:03 this requires yet more software that users have to know to install.

10:06 And also similarly, it doesn't work on much right now.

10:09 Intel says that the binary optimization

10:11 tool currently only works in 12 applications.

10:14 It's also optin, meaning it's not intended to be something you use by default.

10:19 So, we're not testing it by default because it's not

10:22 I they literally say that the 12 applications it supports

10:25 also kind of limits how much we'd be able to see

10:26 an impact anyway because we don't test most of them.

10:28 Assassin's Creed Mirage, Borderlands 3 from 2019,

10:31 Cyberpunk 2077, that one we do test.

10:33 Far Cry 6 2021, Final Fantasy 14 Dawn Trail, we also test Hitman 3 from 2021,

10:39 Hogwarts Legacy, Marvel's Spider-Man Remastered,

10:41 Naraka Blade Point Remnant 2, Shadow of the Tomb Raider from 2018,

10:45 Tiny Tina's Wonderland, and everyone's favorite game, Geekbench.

10:49 Many of these happen to be

10:50 ancient reviewer testing titles seems intentionally selected.

10:54 We've cycled most of them out by now,

10:55 but Cyber Punk and Final Fantasy still overlap for us.

10:58 As for the claim to uplift, onethird of these games get under 3% improvement.

11:02 Half of those are 1 to 2% different,

11:05 which would be indistinguishable from test noise.

11:07 Intel claims 6% in Final Fantasy and 2% in Cyberpunk.

11:11 It's claiming 22% in Shadow of the Tomb Raider,

11:14 which we haven't tested in a long time.

11:16 Now, let's get into some initial data.

11:17 We'll start with frequency validation.

11:19 We perform these tests with logging to ensure the CPU

11:21 is functioning as advertised and as expected from the manufacturer,

11:24 which helps with validation for testing

11:26 and with testing the accuracy of the marketing claims.

11:29 Intel's website claims a 5.3 GHz maximum pcore turbo

11:33 boost for the CPU and 4.6 GHz maximum ecore boost.

11:37 This chart shows the frequency response across all cores in an allcore workload,

11:41 averaging the pores and ecores separately.

11:43 The pecore average held about 5100 MHz during most

11:47 of the test with regular drops to about 5 GHz.

11:50 This is below the 5300 max advertised speed.

11:53 However, the maximum clocks are typically only

11:55 in workloads with the limited thread count.

11:58 The response here is normal given the all core workload.

12:00 The ecore frequency hits and sustains 4.6 GHz.

12:03 Adding the 245k for reference.

12:05 The predecessor averaged 5 GHz even for most of this test.

12:09 The dip down in the center is a logging

12:11 anomaly from the test process and can be ignored.

12:14 In general, the new CPU is 100 to 200 MHz faster.

12:18 Finally, for reference, the 285K hit 5400 MHz during most of this test.

12:22 Its max advertised boost is also higher for single core workloads.

12:26 This chart shows the single core frequency response.

12:28 We're plotting the maximum frequency per interval of all

12:31 cores during a single threaded workload with Cinebench.

12:34 In this task, the CPU technically does

12:36 hit the 5300 MHz advertised boost sometimes.

12:39 But for the bulk of the first part of the workload,

12:41 it's spending more time at 5100 MHz than

12:43 5300 MHz despite being single threaded at this point.

12:47 That's not great.

12:49 The back half of the workload flips that around.

12:51 So, it does eventually spend more time at the advertised 5300 MHz,

12:55 but this at this point is uncommon behavior.

12:57 Adding the 245K to the chart,

12:59 its maximum single thread frequency per interval was

13:01 5200 MHz with regular intervals at 5,000 MHz.

13:05 The 250K Plus is faster by 100 to 300 MHz, but most commonly 100 MHz.

13:11 The 285K's max single thread frequency was 5700 MHz in this chart.

13:15 In BouldersGate 3 at 1080p,

13:17 the Ultra 5250K Plus ran at 107 FPS average with lows at 75 and 66 FPS.

13:24 running with Intel's PP package installed yielded a 109 FPS

13:28 average or a 2% improvement from installing Intel's PP package.

13:32 So, the PP package here isn't doing a whole lot,

13:34 but technically it is getting the job done.

13:37 That's close enough that it may be variance,

13:39 although standard deviation in this title is

13:41 relatively low at 0.95 for average FPS.

13:44 Still, plus or minus one FPS would wash this result,

13:47 so there's no real world impact in this one.

13:49 Lows are also about the same.

13:51 We'll do comparisons against the better of the two for the rest of this.

13:53 109 FPS average with PvP puts Intel's 250 KP just ahead of the 285K nonPP

14:00 with DDR58000 memory and ahead of the like

14:02 forlike 285K's 107 FPS average by 1.7%.

14:07 Against the Ultra 5 245K's 101 FPS average, we're seeing an uplift of about 8%.

14:13 AMD's older AM4 X3D CPUs are still outperforming the 250 KP,

14:18 including the 5700 X3D and the 5600 X3D,

14:21 which is expected to be higher as its clock is higher between the two.

14:25 Both of these are up at 115 to 116 FPS average.

14:28 The older i7 14700 K also leads to 250 KP as does the i5 14600 K,

14:35 although only by 2%.

14:37 The 250 KP is definitely improved on the first revision of Intel's 200 series.

14:41 And at $200, sadly, it's a rare sight these days.

14:44 It's also doing worse than the 14600 K,

14:47 though, and AMD's AM4 X3D lineup remains ahead.

14:50 The 250 KP is substantially cheaper than the 7800 XD and 9800 X3D, of course,

14:55 but is the same price as the launch price for the 5600

14:59 X3D and is $50 cheaper than the original 5700 XD price.

15:03 Not that either can still be found,

15:05 but uh those are good references for where AMD

15:07 has been in this price point in the past.

15:09 Against other CPUs, Intel's 250 KP with its power

15:12 PP outperforms the 9600X's 96 FPS average by about 14%.

15:17 And as you all know, AMD doesn't have a PP package,

15:20 so this is tested without it for the 9600X.

15:22 Outer Worlds 2 is another one of the new

15:24 games we added to our test suite for 2026.

15:27 in this benchmark and at 1080p first.

15:29 The 250K Plus ran at 120 FPS average with PPP and 119 without it.

15:34 So, we're seeing at best a 0.8% improvement.

15:38 In reality, this is likely just margin of error

15:40 and test variance as that falls within those bounds.

15:43 At 120 FPS, the 250KP is about 2 FPS below the 13900K from 2022,

15:49 which itself was about the same as the relatively new 285K.

15:53 Adding faster memory to the 285K didn't

15:55 change much as we've seen in other tests.

15:58 Now to establish the ceiling, the AMD 9850 X2D ran at 143 FPS average

16:02 for a 19.6% advantage on the much cheaper 250 KP.

16:06 For a closer price comparison,

16:08 the 9600X at 96 FPS average was outmatched by the 250 KP by 24%.

16:13 In this one, Intel's 250 KP equaled the performance

16:16 of the 14600K from a few years ago.

16:19 The efficiency may be better,

16:20 but performance hasn't changed much in some of these games.

16:23 As for the 245K, the 250 KP ran 8.2% ahead in average frame rate.

16:28 At 1440p, the 250 KP ran at the same frame rate as at 1080p.

16:32 We're not GPU bound, even at 1440p with our test settings,

16:35 as we've designed the test conditions to lean on the CPU as much as possible.

16:39 In this scenario, there's no meaningful change in the test.

16:42 Even the 9850 X3D actually isn't meaningfully lower than before.

16:45 We're fully CPUbound with the same scaling.

16:47 So, let's move on.

16:48 Stellaris is up now.

16:49 This is tested using simulation time rather than frame rate,

16:52 meaning that there's a real world time difference in how long something takes.

16:56 A faster CPU will allow the game to simulate faster.

16:59 The 250 KP, both with and without PBP,

17:02 took about the same amount of time, both around 39 seconds.

17:06 There's no difference between them.

17:07 We'll take the better of the two results at 39.0 seconds for comparison.

17:11 The 250KP doesn't meet the 285K's performance level in this one,

17:14 with the 285K benefiting from a 1.5%

17:17 simulation time reduction versus the 250 KP.

17:20 That said, the 285K is about $330 to $360 more

17:24 expensive than the 250 KP depending on where you buy it.

17:26 Now, that difference might be worth it in production applications.

17:29 We'll see in a little bit, but it isn't in this situation.

17:32 The 250 KP outperforms AMD's closest modern price comparison, the 9600X,

17:37 by a few fractions of a second in the simulation test.

17:39 These two are about the same in this test.

17:42 The 250 KP out does the 265K by about 1 to two

17:45 seconds and the 5800 XD by about two to three seconds.

17:49 Against the 245K, the 250 KP benefits from a 9.7% reduction in simulation time.

17:55 That's a large improvement for Intel,

17:56 especially since the 245K has also been about $200 lately.

18:00 Just for perspective on this chart,

18:02 the 9850X 2D's results have a time reduction of about 25% from the 250 KP.

18:07 These are totally different by price class,

18:08 but it helps to give perspective to the maximum uplift available.

18:11 We added Kingdom Come Deliverance 2 for our 2026 test suite.

18:15 At 1080p, the 250 KP with PPP ran at 227 FPS average,

18:18 leading the 250 KP without the PP package enabled by about 1.2%.

18:23 The 250 KP is roughly tied with the 285K and follows the 5800 X2D,

18:28 the latter of which has a lead of 3.9% at 236 FPS average.

18:32 People who bought the 5800 XD for gaming and who don't need

18:35 the extra multi-threaded non-gaming performance really

18:38 got out ahead with the 5800 XD, which at this point it's fair to say is goated.

18:42 Against the 9600X for a modern AMD price comparison,

18:46 the 250 KP runs 10% higher frame rate.

18:49 As for the 245K's 206 FPS average, the 250 KP sees a massive improvement of 10%

18:55 while the 245K's launch MSRP was 55% higher.

18:59 These are both moves in the right direction.

19:00 Of course, the 245K wasn't positioned well at launch

19:04 for gaming as Intel's own 14600K outperformed it.

19:07 Previously, we didn't see much benefit

19:09 from faster memory in this test with Intel.

19:12 Our DDR5 8000 C38 kit only boosted the 285K by fractions of a frame per second,

19:18 which is within error and variance, of course.

19:20 At 1440p, the 250 KP ran at 222 FPS average without PvP and 226 with it.

19:26 The stack is overall preserved mostly with the exception of the CPUs at the top.

19:30 We see the 9850 XOD get truncated by about

19:33 20 FPS as a result of the higher resolution.

19:35 There's not much change here since the 250 KP is below the GP's capabilities.

19:40 So, let's continue.

19:41 Dragon Saga 2 is up next.

19:43 In this one, the Ultra 250 KP ran at 106

19:45 FPS average with PPP and about 104 FPS without it.

19:48 They're roughly the same,

19:50 but this matches the trend of slight improvements so far.

19:53 In this case, that's 2.1%.

19:55 The lows also improved, although lows have more variance in the measurements.

19:59 Taking the higher of the numbers, the 250KP ends up just ahead of the 14600K,

20:03 finally beating Intel's prior 5 series CPU

20:06 after the 245K lost to it significantly.

20:09 The 14600K led the 245K by 8% and now the 250

20:14 KP has brought Intel back to where it was in 2023,

20:18 but at least the price improved in a more rapid way.

20:21 The 250 KP also happens to match the 5800

20:24 X2D this time with the 7800 X2D offering minimal uplift.

20:27 The 9800 X2D still has a jump up,

20:30 but the 250 KP is achieving most of the possible performance in this test

20:34 at a fraction of the price of most of the other high performers.

20:37 The 9600X falls behind in this one,

20:40 down at 86 FPS average and yielding a lead of 23% to the 250 KP.

20:45 In Cyberpunk 2077 with 1080p medium settings,

20:48 the 250 KP ran at 169.5 FPS average against

20:51 the 170.2 FPS average results of the version with PvP installed.

20:55 The lows are comparable between the two and functionally the same.

20:58 This is within margin of error and test variance for all three metrics.

21:01 So, we'll use the faster of these to compare.

21:03 As with the others in this game, X2D completely dominates the top of the chart.

21:07 Every X2D CPU except for the 5500 X2D leads here with the 5700 XD,

21:12 5600 XD, and 5800 XD likewise ahead of the 250K Plus.

21:17 The 5800 XD just for reference because so many people have it,

21:20 has a lead of 14% in average frame rate with lows also improved.

21:24 The 5600 XD isn't far behind that and had a more similar MSRP

21:28 for the people who lived near enough to a microenter to buy one at its launch.

21:31 Again, something still sold by AMD and globally.

21:34 The 9600X ran at 158 FPS average and gives up

21:37 a lead of 8% to the 255 KP with PPP.

21:40 The 245K is predictably also beaten by the 250

21:43 KP in this instance by about 4.5%.

21:46 The 14600K is down at 158 FPS average in this game.

21:50 So, this is an instance where it was below the 245K already.

21:53 Increasing the graphics load by moving to high,

21:55 the top end drops from around 230 FPS to 214 FPS average.

21:59 Now for the 9850 X2D, other than the top few,

22:03 the rest of the CPUs aren't affected much

22:04 by imposing more of a bottleneck at the top end.

22:07 The 250 KP with PPP ran at 159 FPS average,

22:10 outdoing the nonPVP test by 1 FPS average.

22:13 Lows are within error of each other.

22:15 The 250 KP is closer to the X2D parts now,

22:18 landing about four to 5 FPS behind the 5600 XD and 5700 XD.

22:23 The 5800 X2D leads by 11%.

22:26 The 9800 XD's lead is less impressive now that it's limited on the GPU load,

22:30 but it's still 31% ahead, which is still pretty good.

22:33 Uh, for a more appropriate comparison,

22:35 the 9600 X ran at 140 FPS average and had lows at 87 and 72 FPS.

22:40 The 250 KP 159 FPS average result is ahead

22:44 of the 9600X by about 14% in this test.

22:47 Just for some references to older parts, users on the 2600, 2700, 3600,

22:52 3700X would see large uplifts by moving to any

22:54 of these newer CPUs from Intel or AMD.

22:57 It's not worth doing that unless you really need the performance.

22:59 Like if you're actively thinking that you

23:02 wish your computer could do something specific better,

23:05 then sure, it might make sense.

23:07 But the big problem, as you all know,

23:09 is that this would require a move to DDR5, and the RAM prices remain crazy.

23:13 So that continues to be the limiting factor for most builds right now.

23:16 But the gains are at least large at this point.

23:18 F125 is up next.

23:20 At 1080p, F-125 positions the 250K Plus with PPP at 253.2 FPS average,

23:26 which is a staggering, unbelievable 0.2 FPS faster than without PPP.

23:34 It truly is incredible what Intel's technology can do.

23:37 Their ingenuity knows no bounds.

23:39 I, for one, am glad that I stayed up all

23:42 night to make sure we got this extra testing in.

23:44 The 250 KP is equal to the 9600X in this game,

23:48 although with slightly better 1% lows.

23:50 Overall, these two CPUs are the same in performance,

23:53 although one of them goes into a socket that has a few more years

23:56 of support and presumably at least one generation after it, and it's not Intel.

24:00 Intel's 250 KP is outdoing the 265K by a few frames per second.

24:05 It doesn't match the 285K in this one, which ran at 267 FPS average,

24:09 and that's both with our standard RAM and with the DDR58000 T38 kit,

24:12 which were tested with an error of each other.

24:14 The 285K with our standard RAM outperforms the 250 KP by 5.8%.

24:19 The 5800 X2D also outdoes the 250 KP,

24:22 but the 250 KP at least runs a higher frame rate than the 5600 X2D,

24:26 although not by much, and then the 5700 X2D.

24:29 The improvement over the 245K's 233 FPS average is 8 and a half%

24:33 for the 250 KP with a similar lead over the 14600K.

24:36 In this one at least, Intel has outdone its predecessor.

24:39 Starfield is up next.

24:40 In this game, the 250K Plus ran at 147 FPS average,

24:43 tying the 265K and also producing better 1% and.1% lows.

24:47 This has the 250 KP just below the 5800 XD.

24:50 Installing PPP improved performance to 150 FPS average about a 2.2%

24:54 2% uplift that allows the 250 KP to marginally surpass the 14600K,

24:59 although they are within error of each other.

25:01 The 285K leads the 250 KP with PPP.

25:04 This time at 2.8% ahead.

25:07 Adding fast memory didn't do much for the 285K in this test.

25:10 In this game, the 5600 XD and 5700 XD

25:12 CPUs don't do as disproportionately well as some other games.

25:16 They're still good, but the 250 KP ends up ahead of them.

25:19 As you'd expect, the 9000 XUD and its refresh end up at the top of the chart,

25:23 but remain in a different price class.

25:24 So, not really an important comparison for this one, just for reference.

25:28 Intel has managed to mostly invalidate its prior 200 series lineup

25:32 with this $200 CPU when it comes to this game at least.

25:35 However, we still have production workloads to look at.

25:38 We're moving on to production tests,

25:39 which is one of the areas that Intel's newer CPUs,

25:41 including the prior 200 series,

25:43 have done a little bit better as compared to their gaming results.

25:46 Extra X3D cache also doesn't offer much

25:48 benefit for AMD in our production test suite.

25:51 So, this can dislodge some of their top

25:53 performing gaming CPUs from the top of the charts.

25:56 In 7zip compression testing,

25:57 we measure performance by calculating how many millions

26:00 of instructions per second each CPU can complete.

26:02 The 250K Plus completed 152.8,000 MIPS with our initial

26:07 run with PVP improving by 1.5% over the original result.

26:11 These results put the 250 KP below the 265K,

26:15 which has a 3.2% advantage over the 250 KPPP.

26:19 It's the quadruple P or the double double P or the double PP if you prefer.

26:24 AMD's 9600X is far down here,

26:27 which isn't news for the six core part in production tests.

26:31 Intel's KP PPP has a 71% lead over the 9600X

26:36 in this compression test and a 1 billion% lead in the letter P.

26:41 Of course, that's the only thing that actually matters.

26:43 Review closed.

26:44 Testing complete.

26:45 You should buy that one.

26:46 This is however from a production standpoint with actual

26:49 performance numbers starting to feel like a mirror

26:51 of 2017 AMD where production was its only real

26:54 bastion of hope before it made gaming work again.

26:57 Maybe Intel's following that road map.

26:59 The 285K at 182,000 MIPS leads the 250 KP

27:04 by 17.7% with AMD's 9950X CPUs at the top.

27:08 As for the 14600 K, Intel's new 250 KP finally manages

27:14 to move the needle beyond what we saw with the 245K,

27:18 which was mostly a disappointing result previously.

27:20 The 250 KP leads the 14600K by 21% in this test.

27:24 Decompression testing is next.

27:26 In this test, the 250K Plus completed 146.8,000 MIPS,

27:31 which matched it with the 9800 X3D.

27:33 The A Core X3D CPUs have never been impressive in this testing,

27:37 as most of their benefit again is in gaming.

27:39 Enabling PVP pushed it up to 151,000 MIPS,

27:42 an improvement from without PvP of 3.2%.

27:46 The 14600K only slightly trails the non- triple P result at 138,000 MIPS.

27:52 The 151,000 MIPS 250 KP result leads the 14600K by about 9% here.

27:58 AMD's 9600X isn't as far down as it was in compression, relatively speaking,

28:03 but the 250 KP still has a 45% lead over the $190 AMD CPU.

28:09 The 285K is about 37% ahead of the 250 KP here.

28:13 So, we're interested to see how the 270 KP does in our next review video.

28:16 Blender is next.

28:17 For this, we run a tilebased CPU render of one

28:20 frame from our logo in the intro animation of these videos,

28:24 but at a higher sample count and quality.

28:26 The 250 KP required 9.2 minutes

28:28 to complete the render under both test conditions,

28:31 that render time requirement has its result reduced from the 14600K by 25.8%.

28:36 With a time requirement reduction against the 245K at 22%.

28:40 The 265K still leads the 250 KP unlike in some

28:43 of the game tests with thanks to its core configuration.

28:46 This is a highly core dependent benchmark.

28:48 So CPUs like the 8 core X2D parts

28:51 fall below performance of higher thread count CPUs.

28:54 This is also why the six core 9600X struggles

28:57 in this one down at 16.4 minutes render time.

29:00 That means the 250 KP required 44% less time than the 9600X.

29:05 Adobe Photoshop testing via the Puget Suite is next.

29:08 This testing measures performance in points

29:10 by running automated sequences of filters,

29:12 scales, warps, and other Photoshop functions.

29:15 Andy's top CPUs here are all functionally the same in performance,

29:19 or they're close enough anyway, with no real advantage between its 16 core

29:22 parts and some of its better eight core parts.

29:25 We don't have the 1400K retested in this one anytime recently,

29:28 and we were too busy benchmarking Crimson Desert

29:31 the past few days to go back for it, but we do have the 14700 K.

29:34 The Fortune 700K has about the same overall score

29:37 as the 250 KP and the R57600 in this test,

29:40 which matches the results we had in this test

29:43 a little over a year ago in terms of relative scaling.

29:45 The 250 KP outdoes the 285K by 4%, including with DDR5 8000 by 1.7%.

29:52 It also bests the 245K by 6.7% below which

29:56 we can find the 5800 X3D and other AM4 CPUs.

29:59 As for AMD's 9600X, which is the main price competition, that's 14% ahead.

30:05 AMD generally does well in Photoshop testing via the Puget Tweet right now,

30:08 which is a flip from several years ago when

30:10 Intel used to be the hold out in this test.

30:12 Chromium code compile testing and Windows is next,

30:15 measured in minutes to build the code base.

30:17 This chart has the 9950 X3D comfortably at the top

30:20 and benefiting from its extra cache and higher core count.

30:23 The first Intel CPU on the chart is the 285K, which credit to it,

30:26 is doing really well here at 106 minutes to complete the compile.

30:30 The 14700K is Intel's next CPU behind that until

30:32 we retest the 14900K in this workload anyway.

30:35 And eventually the 250K Plus, both with and without PPP,

30:38 land at around 133 minutes to complete the compile.

30:41 That means that these entries benefit from a reduction

30:44 in compile time versus the 245K of 21%.

30:47 The 285K and 265K both lead these CPUs though with thanks to the clocks,

30:52 cache, and cores being put to good use here.

30:55 AMD's gaming class CPUs haven't been as competitive

30:57 in this test as Intel in a while now.

30:59 X3D isn't really built for this with a cache helping less

31:02 than more cores or higher clocks on something like the 9800 X3D,

31:07 which is slower than the 250K Plus.

31:08 Other notables include the 9600X down at 235 minutes

31:12 or a full 100 minutes longer to compile than the 250 KP.

31:15 The 9600X is limited in its abilities

31:18 here with its six core 12thread configuration.

31:20 Da Vinci Resolve testing is next.

31:22 We use Premiere internally as we've plugged

31:24 in a bunch of custom automation to it,

31:26 but Resolve is wellresected video editing software for its

31:28 non-subscription approach and its better utilization of hardware for rendering.

31:32 This is an experimental chart.

31:33 We label these clearly to disclose when our confidence

31:36 threshold is lower than it is for other tests.

31:38 We're new to testing Da Vinci Resolve, so to be transparent about it,

31:41 that means there's more risk of an issue with the testing process

31:44 or the data in this chart than there is in other charts.

31:46 We run experimental charts with disclosures because

31:49 we need to begin gaining experience publishing

31:51 it in order to eventually gain the confidence to roll it into the full suite.

31:54 So, we put these experimental charts out

31:56 for the community to offer suggestions for improvement

31:58 so that we can slowly work it

32:00 into the full reviews with our usual high confidence.

32:02 With that clearly disclosed in this one,

32:04 the 250KP scored 12,993 points, putting it below the 265K.

32:09 The 250 KP improves on the 245K by 12% and improves on the 9600X by 16.8%.

32:16 The 5600 XD and 5700 XD give us some pretty valuable information here.

32:20 Because the power budget is the same, but the 5700 XD has two more cores.

32:24 We already know that the 5700 XD

32:26 therefore spreads the same power over more cores,

32:28 resulting in a clock speed deficit as compared to the 5600 XD.

32:32 Because the 5600 XD technically outperforms it here.

32:35 That means that in this test at least,

32:37 Resolve seems to weigh clock speed at least somewhat moderately.

32:40 In other words, it's not as simple as two more cores at lower clocks is better.

32:44 In this case, two fewer cores at higher clocks is better.

32:47 It also appears to make use of the memory change in the 285K

32:50 where we gained 7.4% to 14,383 points from 13,398 points previously.

32:57 This is a much larger gain than we saw in most other tests

33:00 including gaming which typically is more sensitive

33:02 to memory and in particular timings improvement.

33:05 Now for our power test,

33:06 this time since the team was busy on all the case and game testing last week,

33:10 we are keeping it pretty simple.

33:11 For this testing, we set up a PMD2

33:13 as an interposer between the power supply and the test bench.

33:15 So, there's an external capturing device.

33:17 It captures all the power delivered to the bench because the motherboard we're

33:20 using splits some of the 24 pin power to the CPU, which is abnormal.

33:24 We take the ATX 12volt power, the EPS 12volt power,

33:26 add them together, but we subtract out the PCIe slot,

33:29 which we do by isolating the PCIe slot

33:31 on a riser card strictly for this testing.

33:33 So, that card is not there for the performance testing.

33:36 And then we have some overhead for losses

33:38 for VRM efficiency and other 12-volt parts on the board.

33:40 It's not perfect, but it works pretty well.

33:42 This test shows power consumption of the 250

33:44 KP during an allcore Blender workload,

33:46 which will be the worst case power draw scenario for the CPU.

33:50 The external capture utilities using the method described

33:53 have us at about 186 watts for the CPU.

33:55 VRM efficiency losses and miscellaneous board components.

33:58 That's higher than TDP,

33:59 but that's also expected when taking external measurements.

34:02 For reference, a CPU package power measurement using hardware info

34:05 software yields around 140 to 150 watts in the same test.

34:08 We also measured some games as those scenarios tend

34:10 to draw significantly lower power than an allcore workload.

34:13 We'll show just one because it kind of encapsulates all of them.

34:15 At F125 testing, you can see the loading screens

34:18 and cycles with each of these spikes and dips.

34:20 The peaks for externally captured power consumption landed

34:23 at 171 watts with the valleys at 39 watts.

34:26 During low activity during the game itself, such as around the 200 second to 270

34:30 second marks where we're driving around the track,

34:32 we saw around 100 watts from external capture.

34:34 CV package power was closer to 70 watts range via software readings.

34:38 Our traditional power tests are complicated.

34:40 They typically take two days of full time just to process

34:43 the data and arrange it for performance per watt on charge.

34:46 Plus, we need to do about a week of testing to get everything updated.

34:49 So, for now, we'll leave this one here just because of the time,

34:52 but we may revisit it later.

34:53 Overall, compared to Intel's original 200 series launch for the 285,

34:56 265, and 245K CPUs, this is a lot better.

34:59 The original launch was disastrous.

35:01 Total mess.

35:02 The price was just astronomical for some of those CPUs.

35:05 I mean, the 285K one was like $600 plus or minus a little bit.

35:09 That was crazy.

35:11 Not didn't really make sense.

35:12 It was also regressive from the 14th gen.

35:14 So, Intel that launch pretty bad.

35:17 Uh, this one at least looks better.

35:19 It is unfortunately arriving pretty late in the cycle

35:21 for these motherboards in that these are

35:23 probably the last or some of the last CPUs that you'll see in the platform.

35:27 Maybe they'll do one more refresh or something,

35:29 but it looks like this is probably

35:30 the end of the line for the 1851 motherboards.

35:34 And hopefully whatever Intel does next,

35:35 they factor in longevity better because I mean AMD

35:37 basically put out CPUs for AM4 for like 10 years, which is unbelievably cool.

35:42 And that's great for the end user.

35:44 It's uh a lot of people, yes, they like to build a whole new system.

35:48 Every time you build a system, you get a new motherboard, too.

35:50 But especially times like now where whatever the price of the CPU is,

35:55 it almost doesn't matter because the price

35:58 that controls the build is the memory.

36:00 And so in times like now, if you can stay in a platform and kind of like

36:04 nest in it and then just swap CPUs, that's great.

36:07 That's good for everybody except maybe the motherboard makers.

36:10 But ultimately for these new CPUs, the big move here for Intel is the price.

36:13 They did reconfigure them.

36:14 So you've got more cores on the 250KP than you did for say the 245K.

36:18 But it's also cheaper where Intel's 245K was over $300 originally.

36:23 And the 285K, which is now often beaten or tied in gaming with the 250 KP,

36:27 is again $530 to $560.

36:30 So the 250KP being a $200 CPU that's actually reasonable

36:33 in gaming as a standalone port and still pretty damn good,

36:36 actually competitive with AMD,

36:38 better than it in a lot of price for price situations in non-gaming.

36:41 That's good news.

36:42 Intel's 14600K gives it some trouble in some places which does make it

36:45 look a little bad sometimes because now we're getting a couple generations old.

36:48 But in the very least, we're just happy to see literally

36:50 anything even remotely affordable in this market

36:52 because it has been really bad for a long time now.

36:55 And uh you know it's it's tough because the memory is

36:59 still kind of the pricing of it still kind of ruining everything.

37:02 But at like in a total vacuum price

37:06 has come down and their performance has gone up.

37:09 And that's what Intel needs to do if they want to start competing.

37:12 And I am I'm starting to get some of the 2017

37:15 2018 era AMD 10002000 Ryzen vibes from this where Intel

37:22 uh is starting to establish a bit more of a foothold

37:25 at the lower price classes in non-gaming performance in particular.

37:29 It's not to say they're not good in gaming also,

37:31 but uh specifically in production workloads,

37:33 they're starting to establish a foothold there,

37:35 which is how AMD got back into it with the 10002000

37:38 series where you they were kind of okay in gaming,

37:41 but they they really set up their foothold in production.

37:45 And it almost looks like Intel's going that direction right now.

37:47 So, could be pretty interesting in the next few

37:49 years if you can still buy sticks of memory.

37:55 But, but we're reviewing just the CPU today.

37:58 So, a few additional notes.

37:59 The 2DK Plus does have an optional binary optimization tool, but again,

38:04 it only works in two games in our test suite and 12 games in total.

38:08 Of the two in our test suite,

38:09 even Intel itself is only claiming a 2% uplift for one of them.

38:13 After getting all the charts edited and into the review,

38:16 while Vitalia was working on cutting it,

38:18 I did go back through and run the two games for Cyberpunk

38:21 and Final Fantasy 14 with the binary tool uh just to see.

38:25 And I mean, it was what Intel said.

38:27 It was like basically 0% in Cyberpunk

38:29 where the differences were within runto run variance.

38:32 So it's a wash.

38:34 It doesn't do anything in other words or at least nothing that matters.

38:38 And then in Final Fantasy it was like max a couple percentage points.

38:42 So not really a change there for either.

38:44 As for the 200S boost option for say the 285K,

38:47 we ran that CPU back through as well in the final

38:49 stages here of editing just to collect some AB.

38:51 It improves in some cases.

38:53 So if you were to factor this into the earlier charts,

38:55 it'd sometimes shift the 285K to par with a 250 KP rather than a slight loss.

39:00 That said, our methodological or philosophical approach to reviews

39:03 is to define some kind of rigid constraint somewhere because

39:05 otherwise you just get into these uh judgment calls

39:09 of how far do you want to push that arms race?

39:11 Because if you're going to enable stuff that's a separate BIOS setting,

39:14 then you also need to consider that for say AMD where now you need to start

39:17 considering PBO and then to what extent

39:19 do you consider overclocking on the other one.

39:20 So, uh, anyway, we did the numbers at least just to have them.

39:23 But that'll be it for this one.

39:24 Um, yeah, you've got the numbers.

39:26 Hopefully that helps you make a decision if I don't know.

39:30 I I'm really curious like genuinely I am curious how many

39:33 people are in the market to build a computer right now.

39:35 Um, because the disillusionment it doesn't come

39:37 from Intel or AMD uh in the CPU side.

39:39 It really does come from the memory for me where it's just

39:42 every time I look at the RAM prices, I mean, it's bad.

39:44 It's like actually bad.

39:45 It's not sensationalist.

39:46 It's just it is actually just really bad for building a computer.

39:50 Uh, but there's still parts coming out that are interesting.

39:53 So, like I said earlier,

39:54 we're keeping this review really simple by our standards.

39:56 So, a lot of charts there.

39:58 You've got enough to work with.

39:59 The 270 KP will be coming up next for us.

40:02 And I'll keep an eye on comments and stuff to see

40:04 if there's any specific requests for additional testing that we could do.

40:07 Um, but we wanted to just kind of make this one straightforward for us.

40:11 Uh, because I need to sleep and because also just on the team side too,

40:16 everyone's been pushing really hard.

40:17 We've had a killer week for content.

40:18 There's some really cool stuff up there.

40:20 Discord alternatives was awesome.

40:21 The DRAM cartel documentary was really fun to work on.

40:24 Then the knock to a review on the case technical

40:27 side was also we introduced some new technical testing there.

40:30 So, a lot of fun stuff.

40:31 Animation error actually was a big one.

40:32 Simulation time error made its actual real debut

40:35 outside of our white paper piece in Crimson Desert.

40:38 So, that was really exciting, too.

40:39 Anyway, I'm thrilled with the stuff we're putting out.

40:41 But, uh, yeah, I'm going to I'm going to head out and go sleep.

40:44 And 270 KP is next.

40:45 If you have specific requests for additional testing once I

40:48 come back in with more than one hour of sleep,

40:51 then uh we will look at potentially doing that.

40:53 So, let us know.

40:53 But thank you for watching.

40:54 Subscribe for more as always.

40:55 Go to store.gamersex.net to support us directly on patreon.com/gamersex.

40:59 And we will see who next the