News Intel's latest lower-powered CPUs give Ryzen rivals a run for their money — Core i9-14900T beats Ryzen 9 7900 in Geekbench 6 benchmark

[Admin]

The Core i9-14900T has a TDP less than half that of the Core i9-14900K but is only 22% slower in Geekbench 6's multicore test.

Intel's latest lower-powered CPUs give Ryzen rivals a run for their money — Core i9-14900T beats Ryzen 9 7900 in Geekbench 6 benchmark : Read more

 

[bit_user]

In benchmarks and non-gaming usage, the i9-14900T is going to gain a lot from that PL2 of 105 W, which is actually higher than the R9 7900's PPT of 88 W.

Once you hit it with a sustained load (e.g. gaming, rendering, encoding, etc.), you're going to see it throttle back to 35 W, and then you'll really feel how much slower it is at that speed. That's where the 1.1 and 0.8 GHz P-core and E-core base clocks could occur, although hopefully most workloads won't bog that badly.

Furthermore, the default Tau for these CPUs is 28 seconds. So, that's how long it could boost at 105 W. That essentially means you need a half decent cooler on it, or else you're going to hear the fan screaming while doing anything intensive. Since a lot of these are destined for mini-PCs, I'm sure a few buyers with more sensitive ears will be in for a nasty surprise.

 

[Freestyle80]

In benchmarks and non-gaming usage, the i9-14900T is going to gain a lot from that PL2 of 105 W, which is actually higher than the R9 7900's PPT of 88 W.

Once you hit it with a sustained load (e.g. gaming, rendering, encoding, etc.), you're going to see it throttle back to 35 W, and then you'll really feel how much slower it is at that speed. That's where the 1.1 and 0.8 GHz P-core and E-core base clocks could occur, although hopefully most workloads won't bog that badly.

Furthermore, the default Tau for these CPUs is 28 seconds. So, that's how long it could boost at 105 W. That essentially means you need a half decent cooler on it, or else you're going to hear the fan screaming while doing anything intensive. Since a lot of these are destined for mini-PCs, I'm sure a few buyers with more sensitive ears will be in for a nasty surprise.

cant just be happy with competition like a normal person, have to act like a AMD fangirl every comment section

Literally everything you said is ifs buts and maybes, what you doing??

 

[bit_user]

cant just be happy with competition like a normal person, have to act like a AMD fangirl every comment section

I shouldn't dignify this lame ad hom with a response, but I'll take it as an opportunity to explain my position. Not for you, as you've shown yourself to be unworthy of my concern, but in case anyone else is curious.

I just think it's worth highlighting how this CPU has a PL2 that's over 3x its much vaunted TDP, and what that could mean for people. Not only that, but its PL2 is more than 20% higher than the AMD CPU's PPT they're comparing it to.

If you believe in competition, then we need good information about these products, because only when consumers can make choices in their best interest will manufacturers truly be incentivized to build products that best serve their interests.

I don't care if people buy this CPU or not. I just want them to go into that decision with eyes open.

 

[TerryLaze]

In benchmarks and non-gaming usage, the i9-14900T is going to gain a lot from that PL2 of 105 W, which is actually higher than the R9 7900's PPT of 88 W.

Once you hit it with a sustained load (e.g. gaming, rendering, encoding, etc.), you're going to see it throttle back to 35 W, and then you'll really feel how much slower it is at that speed. That's where the 1.1 and 0.8 GHz P-core and E-core base clocks could occur, although hopefully most workloads won't bog that badly.

Furthermore, the default Tau for these CPUs is 28 seconds. So, that's how long it could boost at 105 W. That essentially means you need a half decent cooler on it, or else you're going to hear the fan screaming while doing anything intensive. Since a lot of these are destined for mini-PCs, I'm sure a few buyers with more sensitive ears will be in for a nasty surprise.

While this is for the 12900 they do have the same cores...
These numbers are for groups of 4, the 14900 has 8+16 so 6 groups of 4 total, 6x6 is 36 so at 36W total for the whole CPU is when the e-cores will actually be more efficient, in vector loads, than the p-cores, with the p-cores still contributing.
For integer it would be around 90W

Also from the same link we can see that the clocks this happens at are at above 3Ghz and above 3.5Ghz respectively, far far away from 1Ghz.
So no, it won't be slow, it won't be as fast as the desktop version that is overclocked out of its mind out of the box on almost all mobos, but it won't be slow either.

Also also, tau only works if it is enabled,and also only if it even has thermal headroom to even boost to max, turbo has advanced so much that it will boost as much as it can under any circumstance.

Alder Lake’s Power Efficiency – A Complicated Picture

Reviews across the internet show Alder Lake getting very competitive performance with very high power consumption. For example, Anandtech measured 272 W of package power during a POV-Ray run. Our o…

chipsandcheese.com

With a vectorized workload, Gracemont only beats Golden Cove when running at ultrabook-throttlefest speeds and drawing under 6W.
For integer:
Below 15 watts, Gracemont achieves higher performance while consuming less power than Golden Cove.

 

[TerryLaze]

In benchmarks and non-gaming usage, the i9-14900T is going to gain a lot from that PL2 of 105 W, which is actually higher than the R9 7900's PPT of 88 W.

Furthermore, the default Tau for these CPUs is 28 seconds.

So all of those benchmarks people do in 28 secs...

 

[bit_user]

While this is for the 12900 they do have the same cores...

Not quite. Raptor Lake enlarged larger L2 caches, among other things.

If you want to know how these CPUs perform on a restricted power budget, the best way is to actually measure it. ComputerBase did this with the i9-13900K and found its mulithreaded performance (on their multi-application suite) to be exactly 50% as much, when limited to 45 W, than the default PL2 of 253 W. Using a PL2 of 125 W, as the reference point, 45 W performs 60% as well.

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Leistungsaufnahme und Effizienz

Intel Raptor Lake im Test: Leistungsaufnahme und Effizienz / Leistungsaufnahme in Anwendungen ab Werk / Der Verbrauch steigt

www.computerbase.de

It's a shame they didn't use the 35 W threshold of the T-series, or repeat the testing with the i9-14900K, but we know the i9-13900 and i9-14900 dies are the same stepping, and at least the 45 W figure gives us some real world data to consider.

tau only works if it is enabled,

The only chance you have to disable it is on unlocked, K-series CPUs. On these T-series, the most you can do is to reduce it from the default of 28 seconds (which is counter-productive, from a performance perspective).

For those who don't know, Tau is the maximum amount of time the CPU will boost at PL2, before dropping down to base power (PL1 / TDP). Manufacturers can restrict it beyond the CPU's defaults. On the Dell compact desktop I use at work, they cut it down to just 12 seconds. You can tell it's Tau-limited and not thermally-limited, because thermal-throttling is generally more gradual than Tau-throttling. Also, the CPU temperatures drop under Tau-throttling and yet it still doesn't let the CPU run any faster.

 

[bit_user]

So all of those benchmarks people do in 28 secs...

Probably not, but if a benchmark takes like 45 seconds, then the first 62% of the time elapses before Tau kicks in. So, that boost can still have a marked impact.

 

[DaveLTX]

So a "24 core" CPU that draws quite a bit more power is than a real 12 core?
Surprise surprise!
And in general 7900 isn't that well binned either.

 

[TerryLaze]

If you want to know how these CPUs perform on a restricted power budget, the best way is to actually measure it. ComputerBase did this with the i9-13900K and found its mulithreaded performance (on their multi-application suite) to be exactly 50% as much, when limited to 45 W, than the default PL2 of 253 W. Using a PL2 of 125 W, as the reference point, 45 W performs 60% as well.

Yeah, the same link shows the 7950x at 45W being -77% from the 7950x at 230W...
-50% is still better than -77% .

The only chance you have to disable it is on unlocked, K-series CPUs. On these T-series, the most you can do is to reduce it from the default of 28 seconds (which is counter-productive, from a performance perspective).

No, tau is not part of overclocking, it might have been in the past and it might not be available in laptop bioses, but you can have indefinite TAU on locked CPUs
ALso my point, if you would have read all of it, was more that TAU might not even have been active due to no or not enough cooling.

 

[TerryLaze]

Probably not, but if a benchmark takes like 45 seconds, then the first 62% of the time elapses before Tau kicks in. So, that boost can still have a marked impact.

And that would still be realistic since that's what people would get at home, if you are ok with thinking of no power limit as standard, even though it violates warranty, then you should be ok with this which is even within warranty.

 

[bit_user]

Yeah, the same link shows the 7950x at 45W being -77% from the 7950x at 230W...
-50% is still better than -77% .

Funny thing is, you're the only one talking about a 7950X. The article didn't mention it, and neither did I. I wouldn't recommend that CPU for someone looking to build a 45 W system or below. AMD doesn't even sell a non-X 7950. The highest their 65 W CPU range goes is the 7900.

Injecting an unrelated topic into the discussion is called whataboutism and it's a red flag that someone is posting in bad faith, or is at least working an agenda.

you can have indefinite TAU on locked CPUs

Maybe, but that's news to me. Got proof?

ALso my point, if you would have read all of it, was more that TAU might not even have been active due to no or not enough cooling.

If you're saying the original benchmarks referenced by this article could've hit thermal throttling before Tau, I suppose that's possible. However, that doesn't mean Tau won't hit and Tau-throttling is far worse than thermal throttling (unless you have a truly rubbish heatsink, which it's very unlikely they do).

 

[bit_user]

For sure, the data in the ComputerBase chart is a lot to get your head around. IMO, here's a better way to view it:

​

There are lots of interesting trends you can see. For one thing, performance of the top 3 Ryzens falls sharper below 65 W than Intel. Here, the 7600X is a standout, but then its meager 6 cores are no match for the others.

Also, the Ryzens level off sooner than Intel's Gen 13, which is also hardly news to anyone following this issue. That said, the i5-13600K does get quite flat above 125 W.

What I find notable is that the R9 7900X almost catches the i9-13900K at 65 W and 88 W. I think this gives a pretty clear indication of how much that i9-14900T's scores were helped by its initial turbo boost. In long-running tasks, it would clearly fall well short of a R9 7900 running at 65 W.

 

[TerryLaze]

Funny thing is, you're the only one talking about a 7950X. The article didn't mention it, and neither did I. I wouldn't recommend that CPU for someone looking to build a 45 W system or below. AMD doesn't even sell a non-X 7950. The highest their 65 W CPU range goes is the 7900.

Injecting an unrelated topic into the discussion is called whataboutism and it's a red flag that someone is posting in bad faith, or is at least working an agenda.

Funny how you got an issue with that NOW but not when the original comparison was an 12 core cpu against an 8+16 core CPU...
I didn't see you typing that comparing a CPU with more cores to one with less cores to feed is unfair...
Also the 14900t is just an 14900k limited in power so you have to compare it to the 7950.
Also also the computerbase benchmark THAT YOU BROUGHT UP had the 14900k in it and not the 14900t so what was YOUR point in showing it?!
Also also also I showed that page to show why the 14900t would have good performance at low power, because the e-cores actually become more efficient than the p-cores.

Maybe, but that's news to me. Got proof?

It was a whole thing 2 years ago, Pl1=Pl2=241W ,they stopped using TAU in form of a timed limit, now turbo boost 2 just boosts as far and as long as thermals and average power draw are within the set limits.
Any CPU that has turbo boost 2 has no TAU in seconds anymore.

Intel Turbo Boost 2.0 Technology - SkatterBencher

Intel Turbo Boost 2.0 Technology allows the processor to run faster than the base frequency when it's below its rated specification limits.

skatterbencher.com

 

[TerryLaze]

What I find notable is that the R9 7900X almost catches the i9-13900K at 65 W and 88 W. I think this gives a pretty clear indication of how much that i9-14900T's scores were helped by its initial turbo boost. In long-running tasks, it would clearly fall well short of a R9 7900 running at 65 W.

ZEN 4 cores can't work with low Voltage which means they can't run at low power, the 14900t wins at such low power because of the e-cores.

I mean there is a chance that ZEN 4 improved hugely on that but since they are trying hard to make 4c a thing, I don't think it has.

Zen 2’s only flaw there is its inability to hit very low power levels. Desktop Zen 2 can’t keep scaling down because it hits a voltage floor much earlier than Golden Cove and Gracemont.

Alder Lake’s Power Efficiency – A Complicated Picture

Reviews across the internet show Alder Lake getting very competitive performance with very high power consumption. For example, Anandtech measured 272 W of package power during a POV-Ray run. Our o…

chipsandcheese.com

 

[TerryLaze]

TAU (as seconds) can still be set but is optional and disabled by default.
Data sheet vol.2 page 73.

https://www.intel.com/content/www/us/en/products/docs/processors/core/core-technical-resources.html

The Psys signal and associated power limits / Tau are optional for the system designer and disabled by default.

 

[bit_user]

Funny how you got an issue with that NOW but not when the original comparison was an 12 core cpu against an 8+16 core CPU...
I didn't see you typing that comparing a CPU with more cores to one with less cores to feed is unfair...

The article compared Intel's new flagship low-power socketed CPU with AMD's. As I said, AMD doesn't offer a 7950, at all, so there's no sense in trying to compare against something you can't benchmark or buy.

Trying to turn this into an attack on me is just further evidence that you've lost on the facts and the data.

Also also the computerbase benchmark THAT YOU BROUGHT UP had the 14900k in it and not the 14900t so what was YOUR point in showing it?!

I think it was clear that I cited it to show how the performance of Raptor Lake scales down to lower power levels. I wish I had similar data for the Gen 14 refresh, but they're the same die stepping so it should be nearly identical.

I wasn't actually trying to show anything about Ryzen, but I don't mind a fair look at how their perf/W scaling compares.

It was a whole thing 2 years ago,

Again, please provide evidence that Tau can be disabled for locked and T-series CPUs.

Any CPU that has turbo boost 2 has no TAU in seconds anymore.

Intel Turbo Boost 2.0 Technology - SkatterBencher

Intel Turbo Boost 2.0 Technology allows the processor to run faster than the base frequency when it's below its rated specification limits.

skatterbencher.com

The article clearly states they're referring to K-series CPUs.

The slide says "unlocked"! Do you look at this stuff before posting???

 

[bit_user]

ZEN 4 cores can't work with low Voltage which means they can't run at low power, the 14900t wins at such low power because of the e-cores.

I mean there is a chance that ZEN 4 improved hugely on that but since they are trying hard to make 4c a thing, I don't think it has.

Alder Lake’s Power Efficiency – A Complicated Picture

Reviews across the internet show Alder Lake getting very competitive performance with very high power consumption. For example, Anandtech measured 272 W of package power during a POV-Ray run. Our o…

chipsandcheese.com

Those charts show Zen 2. Try harder.

 

[bit_user]

TAU (as seconds) can still be set but is optional and disabled by default.
Data sheet vol.2 page 73.

https://www.intel.com/content/www/us/en/products/docs/processors/core/core-technical-resources.html

That's funny: when I look at the datasheet (vol 2) for Gen 13 CPUs, page 73 is talking about Programmable Attribute Maps for memory segments. A search of that entire document doesn't even mention "tau"!

Are you sure you're not just hallucinating?

 

[TerryLaze]

That's funny: when I look at the datasheet (vol 2) for Gen 13 CPUs, page 73 is talking about Programmable Attribute Maps for memory segments. A search of that entire document doesn't even mention "tau"!

Are you sure you're not just hallucinating?

You got me,it was on volume 1 and they moved it around and it is now on page 78.

Again, please provide evidence that Tau can be disabled for locked and T-series CPUs.

Title of the document, it's for all the CPUs.

 

[TerryLaze]

Those charts show Zen 2. Try harder.

Try harder at what? Computerbase is for zen 4 and they show better performance on 13900k at low power than they do for zen 4, and computerbase is with locked power levels so no TAU even if it were still a thing, and they use benchmarks that take a good long while, especially at low power.
So again what should I even try?!

 

[bit_user]

Computerbase is for zen 4 and they show better performance on 13900k at low power than they do for zen 4,

Yes, I said as much.

However, the Toms article isn't comparing a R9 7900 at 45W, it's comparing one at its default power setting. Using the data from ComputerBase, you can clearly see how a R9 7900 would actually perform at 65 W vs. a Raptor Lake i9 at < 45 W.

computerbase is with locked power levels so no TAU even if it were still a thing,

Yes, there's no Tau, but also no boost for the Ryzens. My original post was about the implications of the large turbo power limit and the performance drop experienced, once you hit Tau and have to run at PL1. The ComputerBase article shows the sustained performance levels.

You got me,it was on volume 1 and they moved it around and it is now on page 78.

It's not a user manual. They're describing the behavior of these parameters for system designers to use.

It sounds to me like the effect of Tau=0, at this level, is to disable Turbo Boost 2.0 and then it just always runs according to PL1.

Title of the document, it's for all the CPUs.

Page 87 states:

"May vary based on SKU."
​

 

[TerryLaze]

Yes, I said as much.

However, the Toms article isn't comparing a R9 7900 at 45W, it's comparing one at its default power setting. Using the data from ComputerBase, you can clearly see how a R9 7900 would actually perform at 65 W vs. a Raptor Lake i9 at < 45 W.

That has nothing to do with the fact that zen 4 drops like a stone at low power, and according to chipsandcheese it's because they can't work on low voltage.
Unless you can find some proof that they fixed the low voltage issue on later gens.

Yes, there's no Tau, but also no boost for the Ryzens. My original post was about the implications of the large turbo power limit and the performance drop experienced, once you hit Tau and have to run at PL1. The ComputerBase article shows the sustained performance levels.

All of them run without any boost, that's the whole point of doing these kind of benches in the first place.Even if the geekbench scores are "not real" or were running under turbo, Conputerbase runs them like for like, low power, 13th gen still better than zen 4, for the normal parts at least but for intel the t versions are the same as the k just with (different) power limits, if the non-x parts are different, other than clocks, tell us.

 

[bit_user]

That has nothing to do with the fact that zen 4 drops like a stone at low power, and according to chipsandcheese it's because they can't work on low voltage.

Again, you're confusing or conflating Zen 2 and Zen 4. Also, more whataboutism.

All of them run without any boost, that's the whole point of doing these kind of benches in the first place.Even if the geekbench scores are "not real" or were running under turbo, Conputerbase runs them like for like, low power, 13th gen still better than zen 4, for the normal parts at least but for intel the t versions are the same as the k just with (different) power limits, if the non-x parts are different, other than clocks, tell us.

I never said GeekBench weren't real or realistic. Just that they don't capture all real-world usage scenarios, like gaming and other tasks involving sustained loads.

As for the ComputeBase data, I've said all I have to say about that. I think this plot speaks volumes:

​

 

[prollie]

It's great to see we're getting more solid performance out of low power desktop chips. Price vs performance does leave a lot to be desired from Intel's side, but none the less the ball *is* rolling.

IMO Intel's price disadvantage really isn't just a technicality with this one though, it's a hard issue. I think it's reasonable to assume street price will add another $25-30 ish on Intel's 1000+ tray price.

Looking at Am4zon right now, this means for the AMD option you can get a whole AM5/DDR5 platform, *at the start* of its lifecycle, for the same price: An R9 7900 w/cooler, a B650 board and a 32gig DDR5 6000mhz kit - for the price of just the 14900T chip. Heck, even if you already have some spare DDR4 and thus "save" yourself an additional ~$65 expense, you still gotta tack on the price of an LGA1700 compatible cooler and an LGA1700 board to have a working platform. An already "EOL" platform for all 3 components, possibly even the cooler too.

That makes the 14900T proposition a *hella* rough one. The chip itself in a vacuum, it's performance does have good merits. But as a real-world product option it's downright awful value.