News AMD beats Intel in mobile gaming despite new and refreshed CPUs — Ryzen 9 7945HX3D prevails over the Core i9-14900HX in new review

[Admin]

Notebookcheck recently reviewed Intel's new i9-14900HX mobile flagship and found in real-world testing that its performance is lackluster, particularly in gaming. It failed to outperform its 13th-gen counterparts outright and could not beat AMD's Ryzen 9 7945HX3D in gaming-centric workloads.

AMD beats Intel in mobile gaming despite new and refreshed CPUs — Ryzen 9 7945HX3D prevails over the Core i9-14900HX in new review : Read more

 

[usertests]

This is where desktop X3D shines, efficiency.

Meteor Lake iGPU vs. Phoenix/Hawk iGPU is the real fight.

 

[Alvar "Miles" Udell]

Notebookcheck's review reveals that Intel's new Core i9-14900HX is an underwhelming chip, providing lackluster performance.

I think you meant to say "Lackuster performance improvement over the 13900HX" because it doesn't have lackuster performance.

But it doesn't take a review to see it doesn't have much different than the 13900HX, a glance at the specs from Intel show the only difference is a 200mhz higher E core boost speed and 300mhz higher P core boost speed vs the 13900HX.

The real question is what will 14900HX laptops cost vs their identically spec'd 13900HX models.

 

[Sleepy_Hollowed]

Sounds about right.

Intel was asleep on the iGPU for the longest time, serves them well, especially after laying off not just from sales and other positions, but they just did not care about R&D for the longest time.

 

[Amdlova]

Beat on what? It's a nightmare to find a powered device with amd. You can find intel notebooks at car shops. The last amd notebook I see need to be imported from united states... Scarce product make amd fell so premium. Premium to the wallet...

 

[MXM0]

Not sure how I should feel about an Intel 7 chip being only 6% less performance than a TSMC N4 (or is it N3?) chip, when Intel 3 and 20A are right around the corner. If there wasn't a huge power difference, then AMD would have a really, really bad problem.

 

[Moobear]

Not sure how I should feel about an Intel 7 chip being only 6% less performance than a TSMC N4 (or is it N3?) chip, when Intel 3 and 20A are right around the corner. If there wasn't a huge power difference, then AMD would have a really, really bad problem.

Not going out much? The X3D are older hardware compared to the 14 series this year the new amd cpu s come out and guess how they are going to look in benchmarks.

 

[George³]

feel about an Intel 7

Put away your humiliated feelings now. Intel is losing, the facts are. 😉

 

[fret]

Not sure how I should feel about an Intel 7 chip being only 6% less performance than a TSMC N4 (or is it N3?) chip, when Intel 3 and 20A are right around the corner. If there wasn't a huge power difference, then AMD would have a really, really bad problem.

Agreed. NotebookCheck reviewed a reference device in which we don't even know what the PL1 limit is,

The AMD is 5nm chip.

 

[fret]

So, per Notebookcheck, last years Intel 13000HX had a Cinebench Multi Core Score RANGE between: 24,270 and 32,623. That's a huge gap in performance right.

Well, not all laptops are the same; some are thinner, some are thicker; some are meant to be cooler and quite, some hotter and louder.

Well, per Notebookcheck, this new chip gets a score of: 23,336 in this one lone Engineer's laptop. As of right now, that is a worse score than all of last years 13th gen, 13900HX laptops that Notebookcheck have tested thus far.

Grain of salt needs to be taken with these scores. It is a Reference / Engineers device after all.

 

[cyrusfox]

What in the world is new about the 14900HX... This is old news.

I only saw the first part of the headline and though this was going to be about Meteor lake, like actual results showing Intel losing, rather than talking about the HX parts which are more often than not paired with a dGPU any ways... The HX series is the biggest mobile SKU given Desktop power envelope to spread its wings, the iGPU is only there for media and other task so the dGPU does not kill the battery.

Grain of salt needs to be taken with these scores. It is a Reference / Engineers device after all.

Its all based how well they are designed thermally, if it can sustain higher power draw, its a desktop class performant chip.

 

[fret]

So, per Notebookcheck, last years Intel 13000HX had a Cinebench Multi Core Score RANGE between: 24,270 and 32,623. That's a huge gap in performance right.

Well, not all laptops are the same; some are thinner, some are thicker; some are meant to be cooler and quite, some hotter and louder.

Well, per Notebookcheck, this new chip gets a score of: 23,336 in this one lone Engineer's laptop. As of right now, that is a worse score than all of last years 13th gen, 13900HX laptops that Notebookcheck have tested thus far.

Grain of salt needs to be taken with these scores. It is a Reference / Engineers device after all.

PS: Kind of sad that I'm the one who has let this be known.

But unfortunately, to anyone who has read this article, the damage as already been down. Most who read this article will already assume the 14 Gen HX chips are garbage.

Mission accomplished?

 

[usertests]

The HX series is the biggest mobile SKU given Desktop power envelope to spread its wings

It's actually desktop silicon inserted into a mobile package, so it has 8 P-cores instead of 6 and so on.

AMD's own take on "HX" is a little confusing because for 5000-series (e.g. 5980HX) and 6000-series (e.g. 6980HX), they were based on the Cezanne and Rembrandt mobile dies. Then it became 7045-series "Dragon Range" (e.g. the 7745HX and 7945HX3D), desktop Raphael put into mobile packages like Intel was already doing. They will also change the '5' to a '0' for a lower clocked version.

 

[Avro Arrow]

"AMD's Ryzen 9 7945HX3D counterpart was far superior in the same area — sporting slightly higher FPS while consuming far less power."

For mobile parts, substandard power efficiency is the kiss of death because that directly affects the battery life. Notebookcheck's review is kinda weird though because it doesn't give actual power draw numbers when not at idle, it just has this weird "points" system. Looking at this crazy power-over-time chart for Cinebench R23 multi-core, I took each of the seven verticies, added the watts together and divided by 7 to get an approximate average of 158W over the 20-second period (for all the good that does):

 

[fret]

"AMD's Ryzen 9 7945HX3D counterpart was far superior in the same area — sporting slightly higher FPS while consuming far less power."

For mobile parts, substandard power efficiency is the kiss of death because that directly affects the battery life. Notebookcheck's review is kinda weird though because it doesn't give actual power draw numbers when not at idle, it just has this weird "points" system. Looking at this crazy power-over-time chart for Cinebench R23 multi-core, I took each of the seven verticies, added the watts together and divided by 7 to get an approximate average of 158W over the 20-second period (for all the good that does)

Plus there is this: From the Notebookcheck review:

"Please note: During our power measurements using an external display, the dedicated graphics card was active—this was also the case on quite a few of the comparison systems. The video output's configuration largely depends on the laptop manufacturer's configuration, the XMG Neo 17 (E24)'s HDMI output and its USB-C port are connected directly to the dGPU."

My Gaming Laptop also uses the Dedicated Graphics when using an External Display, which ruses a lot of watts; which is somewhat why I hate when Benchmarks are run "Off Screen" or to an "External Display". Doing such can further muddy up the waters when comparing total watts being used.)

 

[TerryLaze]

"AMD's Ryzen 9 7945HX3D counterpart was far superior in the same area — sporting slightly higher FPS while consuming far less power."

For mobile parts, substandard power efficiency is the kiss of death because that directly affects the battery life. Notebookcheck's review is kinda weird though because it doesn't give actual power draw numbers when not at idle, it just has this weird "points" system. Looking at this crazy power-over-time chart for Cinebench R23 multi-core, I took each of the seven verticies, added the watts together and divided by 7 to get an approximate average of 158W over the 20-second period (for all the good that does):

So they tested power for 20 secs on a CPU that has 28 secs of turbo boost...makes sense.

 

[Avro Arrow]

Plus there is this: From the Notebookcheck review:

"Please note: During our power measurements using an external display, the dedicated graphics card was active—this was also the case on quite a few of the comparison systems. The video output's configuration largely depends on the laptop manufacturer's configuration, the XMG Neo 17 (E24)'s HDMI output and its USB-C port are connected directly to the dGPU."

My Gaming Laptop also uses the Dedicated Graphics when using an External Display, which ruses a lot of watts; which is somewhat why I hate when Benchmarks are run "Off Screen" or to an "External Display". Doing such can further muddy up the waters when comparing total watts being used.)

Absolutely. They should only test models that don't have dGPUs. I've never seen a craptop that doesn't have an IGP so it shouldn't be too hard. OTOH, if they both have the exact same dGPU then it would still be valid. Craptops are tough to gauge because they all have different specs, different batteries, etc. that make an apples-to-apples test almost impossible.

This is why whenever I'm in the market for one, I set a price that I'm willing to pay and just get the best for that price. My current model is an ASUS Vivobook with an AMD R5-3500U and I chose it over the other craptops at the same price point simply because it has a (mostly useless) GTX 1050M dGPU and a backlit keyboard. Since I don't use craptops for anything strenuous (I have my desktop monster for that), they tend to live a long time with me as their owner.

Hell, I recently resurrected my 8 year-old Acer Aspire with an A8-3500M and Windows 7 to replace an equally old Android tablet that finally became too slow to be useful. It's actually better at the job that the tablet had, that of a bathroom reader! 🤣

 

[fret]

So they tested power for 20 secs on a CPU that has 28 secs of turbo boost...makes sense.

Trying to determine a CPU’s Efficiency while a CPU is using Watts beyond the Designed TDP of the Chip (which is called Turboing) is pointless. Turboing is Terribly Inefficient.

Here is an example. I’m going to compare an Intel Ultra 7 155H to ……..itself.

Cinebench R23 Multicore scores:

Intel 155H with a True TPD of 28 watts but running a Sustained wattage of 45 watts (Turboing) = 14,500.

Now, the same chip running at it’s True TDP of 28 Watts = 13,000.

So, lets do math. It takes 60.7 % more energy (45/28 = 1.607) to get only 11.5% more performance (14,500/13,000 = 1.115). See, Turboing is Terribly Bad.

So, the same chip gets either: 322 points per watt (14,500 / 45 watts) or 464 points per watt (13,000 / 28 watts) depending on if its Turboing or not.

Apple's M Series of Chips tend to be known for the best Performance / Watt. However, they never Turbo. Their Max Sustainable Performance is maintained at their Designated TDP.

To compare further, An M2 Pro 10-Core with a TDP of 27 watts (and running at a Sustained 27 watts) gets a score of 11,800. 437 points per Watt (11800 / 27)

What is likely to be surprising to over 90% of the people out there, the Intel 155H is more Efficient than an M2 Pro running at the same TPD (when both chips are NOT Turboing). 13,000 vs 11,800 at their Designed TPD’s of 28/27 watts.

What is frustrating, is when people see the Intel Ultra 7 155H get a score of 14,500, most people think it is using 115 watts to get that score because Intel has "115 Watts" as this chips "Max Turbo Power" watts. (Intel's "Max Watts" number is the CPU watt PLUS the GPU watts both running at MAX at the same time; not the CPU alone.)

The saddest part is that Tech Site never explain this stuff. UGH.

 

[fret]

PS: An M3 Pro with a TDP of 27 Watts and running at a Sustained 27 watts gets a Cinebench R23 Score of 15,000.

So, lets look a little closer at 3 chips all running at their TDP's of 27/28 watts:

M2 Pro: 11,800 (5nm chip)
Intel Ultra 7 155H: 13,000 (4nm chip)
M3 Pro: 15,000 (3nm chip)

See how the Intel Ultra score sits right in between the 2 Apple chips scores? Well, it should. 4nm sits somewhere between 5nm and 3nm. Hence then name Intel 4.

 

[fret]

So, back to the main subject:

If you've read my above posts, and have an understanding of what I said, you should now NOT be surprised that a 55 watt TDP, 5nm AMD is better than a 55watt TDP, 7nm Intel chip.

 

[TerryLaze]

Trying to determine a CPU’s Efficiency while a CPU is using Watts beyond the Designed TDP of the Chip (which is called Turboing) is pointless. Turboing is Terribly Inefficient.

Yes, that's my point, I was being sarcastic pointing out that the power test was conducted only during the turbo time.

Although turbo is within the TDP.

 

[MXM0]

Put away your humiliated feelings now. Intel is losing, the facts are. 😉

I'm not sure I understand what you mean. If this is about user perception of "AMD vs Intel", AMD is looking like they're at a dead end for the time being, unless Intel hands them space on their production lines, and access to PowerVia? This article is about notebook CPUs, but.. if I were building an upgradable gaming rig in the near future, AMD would probably be a poor choice (I do have a 5000 series v-cache now, and I love it, for ref).

Now that userbenchmarks has picked up some builds with the 14900HX, it looks like its identical perf to the AMD chip, but $100 cheaper. It uses more power, yes, but.. it's on a larger node. The sad part is that it doesn't have anywhere near the transistor count of the 3D cache chip, and it still matches perf. It's power hunger comes from the larger process, plain and simple.

Given that TSMC is reported to not be interested in buying any more EUV machines from ASML until 2030 (they certainly aren't this year -- Intel has 60% of the machines total for 2024, and TSMC has none), it's looking like Intel will catch up, and pass, in really short order.

 

[fret]

Yes, that's my point, I was being sarcastic pointing out that the power test was conducted only during the turbo time.

Although turbo is within the TDP.

The TDP for this chip is 55 Watts (Base Power). Efficiency after this point begins to fall. (note, it is not in a linear fashion either).

The "Max Turbo Power Value" is the most the ENTIRE Processor can handle, including the iGPU when the CPU and iGPU are running at the same time.

Statement from Intel's Website:

"Is power consumption measured separately for the processor and the integrated graphics controller?
TDP is calculated assuming core and graphics combined, but it's also reported separately for the core and the graphics."

"How can I check the TDP of my Intel processor?

• Go to product specification site (ARK).
• Enter your processor number.
• Under Performancesection:
  • Check the value of the TDP field for processors older than the 12th Gen Intel® Core® Processors.
  • Check the value of the Processor Base Power field for 12th Gen Intel® Core® Processors and above."

Notice how they say the "BASE POWER" is the TDP. Which for this chip is 55 watts. Sure, the Entire Processor can handle more; (and at a huge cost of Efficiency, because anything after that is Turboing)

Here's the source: https://www.intel.com/content/www/us/en/support/articles/000055611/processors.html?wapkw=tdp

 

[fret]

Yes, that's my point, I was being sarcastic pointing out that the power test was conducted only during the turbo time.

Although turbo is within the TDP.

PS: Please don't think I'm being hateful or such to you. I'm not intending to be. Sorry if it seems that way.

In fact, I just learned most of this stuff myself.

The problem is that Tech Sites do not explain this stuff to their Audience, and therefore can cause a lot of misconceptions / understanding.

I began researching on my own after following discussions / arguments in forums. Heck, I'm sure I made an ass out of myself in some of those discussions too.

It's just so frustrating that such complex things like computers and such are not dumbed down to us common folks.

UGH.

 

[fret]

I'm not sure I understand what you mean. If this is about user perception of "AMD vs Intel", AMD is looking like they're at a dead end for the time being, unless Intel hands them space on their production lines, and access to PowerVia? This article is about notebook CPUs, but.. if I were building an upgradable gaming rig in the near future, AMD would probably be a poor choice (I do have a 5000 series v-cache now, and I love it, for ref).

Now that userbenchmarks has picked up some builds with the 14900HX, it looks like its identical perf to the AMD chip, but $100 cheaper. It uses more power, yes, but.. it's on a larger node. The sad part is that it doesn't have anywhere near the transistor count of the 3D cache chip, and it still matches perf. It's power hunger comes from the larger process, plain and simple.

Given that TSMC is reported to not be interested in buying any more EUV machines from ASML until 2030 (they certainly aren't this year -- Intel has 60% of the machines total for 2024, and TSMC has none), it's looking like Intel will catch up, and pass, in really short order.

We'll see how it all goes, but I feel the same as you do.

If everything that has been reported is True, Intel is going to leapfrog TSMC in quick order. (Unless TSCM has some unforeseen Trick Up their Sleeve)