Intel Core i9-13900K

[Gameslove]

Big big disappointment: a lot of powers consumption and very high temperature.

Not full gaming parity vs Ryzen 7 5800x3d (according all reviews) .

Overall a Ryzen 7 5800x3d stay in 1 place efficient / gaming.

 

[ShiningSapphire]

Other reviewers are seeing higher results in CB23 with power limits removed.

Something very wrong must be with their testing methodology, it was the same with the 4090 tests - much lower results compared to other reviewers because they were testing on a 5800X which was an obvious bottleneck. The worst of this is probably the temperature tests, fanboys everywhere are screaming about 117 degrees but they don't see or don't want to see that the measurements were taken on a tiny Noctuy air cooler. LOL
By the way, RPLs have great potential for undervolting.

 

[R0H1T]

The U14s isn't tiny, in fact the test shows that you will need to invest a lot more in cooling the 13xxx chips if you want them to stay reasonably cool even with extreme workloads. More than AMD I'd say, so there's the extra investment needed for that.

 

[fevgatos]

Dude. I'm not saying anything about AMD products. Wrong thread though. I was point something out about your comments to my post.
I honestly don't care about your glaring problems with Intel's praise and my comments are not for anyone's amusement.


Im not comparing anything. I just expressed something about a product that just came out.

Ιm just asking, in your opinion, how faster is the 7950x over the 12900k. Gimme a straight answer if possible

The U14s isn't tiny, in fact the test shows that you will need to invest a lot more in cooling the 13xxx chips if you want them to stay reasonably cool even with extreme workloads. More than AMD I'd say, so there's the extra investment needed for that.

Well I have a u12a, pretty similar to the u14s, my 12900k is sitting at 78c in CBR23 at stock.... This site has it at 95 or something.

 

[R0H1T]

You're not running it on manufacturer (board) defaults though are you?

 

[HTC]

The worst of this is probably the temperature tests, fanboys everywhere are screaming about 117 degrees but they don't see or don't want to see that the measurements were taken on a tiny Noctuy air cooler.

The U14s isn't tiny

It's actually a big cooler, though there are bigger air coolers.

@W1zzard

Could you perhaps make a thermal throttling test for this CPU like you did with the 7950X?

 

[ShiningSapphire]

The U14s isn't tiny, in fact the test shows that you will need to invest a lot more in cooling the 13xxx chips if you want them to stay reasonably cool even with extreme workloads. More than AMD I'd say, so there's the extra investment needed for that.

There is nothing new about the fact that such cooling is not enough for such a powerful cpu. Who want to buy this will of course go for a 360/420mm AiO so what's the point of this testing procedure, to show that the processor will boil under poor cooling? This is obvious.

 

[R0H1T]

Who want to buy this will of course go for a 360/420mm AiO so what's the point of this testing procedure, to show that the processor will boil under poor cooling? This is obvious.

You don't speak for thousands of others who will buy these chips, in fact many on this forum prefer air cooling. Besides not every case will have space for 360/420mm cooling, oh wait is Intel gonna subsidize that as well through their contra revenues "marketing development funds"

 

[HTC]

here is nothing new about the fact that such cooling is not enough for such a powerful cpu.

Are you sure about that? A LOT of people were claiming a "heavy-duty cooler" was required for a 7950X (it's most direct competitor) but @W1zzard proved that it could be run just fine, even with a Wraith cooler using fan @ 20%, though @ the cost of A LOT of performance.

Would the 13900K behave the same way? I don't know.

 

[Chrispy_]

From my cristal ball: 65w 13300f\13100f on ddr4+h610 will be the ultimate budget options with better gaming preformance all around. AM5 on ddr5 just can't cut it unless massively priceed cut.

AM5 is probably a little premature as a DDR5-only platform. The 13100F on an mATX H610 board will almost certainly be great bang for the buck at the very lowest end.

I think the new 13th gen label slapped on the rebranded 12400F isn't going to change anything, it'll still compete with AM4, and a cheap B550 board and DDR4-3600 kit pairs exceptionally well with a 5700X. There are no winners or losers in that fight, both platforms are a dead end, both are good performance/$ and both are competent, all-round solutions that will likely outsell higher-tier products from either brand by one or more orders of magnitude.

 

[fevgatos]

You're not running it on manufacturer (board) defaults though are you?

Yes I am

 

[Chrispy_]

Are you sure about that? A LOT of people were claiming a "heavy-duty cooler" was required for a 7950X (it's most direct competitor) but @W1zzard proved that it could be run just fine, even with a Wraith cooler using fan @ 20%, though @ the cost of A LOT of performance.

Would the 13900K behave the same way? I don't know.

No need to wonder. HUB did the testing and it's ugly for Raptor lake.



Yes, the 7950X also starts to suffer right down at 65W, but the difference at 105W is insane, with it outperforming the 13900K by over 50%.

Edit:
These results have been retracted by HUB, though IMO they are very much still valid, although misleading. The terrible showing above is a direct result of Intel's own XTU software which is not something you'd get if you power-gated Raptor lake in the BIOS. Blame Intel for shit software because that's the true source of this erroneous data.

 

[HTC]

No need to wonder. HUB did the testing and it's ugly for Raptor Lake at 65W:

View attachment 266536

I meant using the same kind of approach as in that 7950X thermal throttling test: not by reducing the power, but rather by reducing the fan speed to simulate different (worse) coolers, in order to test how the CPU would fare with them as well as how much performance it would lose with the "worse coolers".

 

[Solid State Brain]

No need to wonder. HUB did the testing and it's ugly for Raptor lake.

View attachment 266536

Yes, the 7950X also starts to suffer right down at 65W, but the difference at 105W is insane, with it outperforming the 13900K by over 50%.

The HardwareUnboxed results have been retracted:

https://twitter.com/i/web/status/1583257536202842114

 

[Dirt Chip]

What you save today by not buying motherboard if you come from alder then that same you save when ryzen 5 comes. So either way gets to the same result more less.

You mean use the same 6xx mobo with zen5?
If you are on a budget it will be very poor choice, budget wise, to change CPU every year or two.
Also, most people dont change cpu often, and some of one who do tend to upgrade the mobo as well as cpu to get all the new and fresh tech.
The whole "dead platform" seems irrelevant to me (I upgrade every 10 years or so on average), but I can see why on forums like this is quite prevalence (although I don't think it's the common state of mind).

 

[Chrispy_]

The HardwareUnboxed results have been retracted:

Oh, good spot.
I guess we'll have to wait for W1zzard to have time for a power-scaling test of his own.

 

[R0H1T]

The whole "dead platform" seems irrelevant to me (I upgrade every 10 years or so on average), but I can see why on forums like this is quite prevalence (although I don't think it's the common state of mind).

It is important in more ways than one, chances are if your mobo dies for some reason you'll be able to buy a cheap second hand or even brand new x570 board 2-3 years down the line. Good luck finding a z97 board(used or new) though at reasonable prices. If you want to upgrade you could go from zen to zen3 as well with the chipset limitations, the last time Intel allowed this was when Youtube was barely a thing. Intel's artificial limitations wrt sockets are well documented & really there's no excuse for changing them 5 times on the same uarch, with Skylake!

This is even more relevant now because PCIe 5.0 will easily last you a decade or more, we have no dGPU's which can make use of it & barely any SSD's in the consumer space which can properly use it. This wasn't the case a decade back, because while the progress from PCIe 2.0-> 3.0-> 4.0 was painfully slow the jump from 4.0->5.0-> 6.0 will take less than half the time. So in essence these boards will last you for a long while!

 

[Chrispy_]

It is important in more ways than one, chances are if your mobo dies for some reason you'll be able to buy a cheap second hand or even brand new x570 board 2-3 years down the line. Good luck finding a z97 board(used or new) though at reasonable prices. If you want to upgrade you could go from zen to zen3 as well with the chipset limitations, the last time Intel allowed this was when Youtube was barely a thing. Intel's artificial limitations wrt sockets are well documented & really there's no excuse for changing them 5 times on the same uarch, with Skylake!

This is even more relevant now because PCIe 5.0 will easily last you a decade or more, we have no dGPU's which can make use of it & barely any SSD's in the consumer space which can properly use it. This wasn't the case a decade back, because while the progress from PCIe 2.0-> 3.0-> 4.0 was painfully slow the jump from 4.0->5.0-> 6.0 will take less than half the time. So in essence these boards will last you for a long while!

Case in point, I sold an unopened X99 board a couple of months back for twice what I originally bought it for, becuase S2011-3 CPU users are stiffed if their board dies. Nothing else is compatible.

I was on the other end of the same situation where I was putting together a build to gift someone and intended to repurpose an i9-10900 I had. Could I find a decent S1200 board new for any reasonable price? No. I ended up buying a used Asus Prime Z490 board for about what they were selling for brand new three years ago, dusty, unknown history, no accessories other than the IO shield and absolutely no warranty.

The previous machine I built at home as part of my regular parts cleanouts of abandoned crap that ends up at home used a Ryzen 5 1600 and I had a B450 board lying around, but I can still buy cheap mATX B550 and A520 boards for next to nothing, brand new, with a warranty that will work just fine with a CPU that's two years older than the Comet lake i9 I struggled to find a board for last week.

 

[R0H1T]

Yes this planned obsolescence should not be condoned by anyone, it's really only good great for Intel & their board partners & horrible for the end user! AMD's done some bad things especially wrt pricing their 5xxx or 7xxx chips but socket longevity is not one of them, it's at least +10 IMO over any other feature Intel can show in good light with their 13xxx launch.

 

[AnotherReader]

You mean use the same 6xx mobo with zen5?
If you are on a budget it will be very poor choice, budget wise, to change CPU every year or two.
Also, most people dont change cpu often, and some of one who do tend to upgrade the mobo as well as cpu to get all the new and fresh tech.
The whole "dead platform" seems irrelevant to me (I upgrade every 10 years or so on average), but I can see why on forums like this is quite prevalence (although I don't think it's the common state of mind).

I've done that, and it's a valid strategy. Don't forget that you can sell your previous CPU. I bought a 1700X and X370 Taichi in 2017 only a couple of months after release. I swapped the CPU for a 3600X that ended up costing me only $75 and came with a free game as well. Now I have a 5700X and my 3600X should be in someone else's hands soon. The total outlay is slightly more than buying an 1800X in 2017, but a 5700X is much faster. Alternatively, you can upgrade at the end and you only spend on the CPU rather than CPU and motherboard.

 

[Why_Me]

 

[QUANTUMPHYSICS]

My flight simulator is using an older Core i7 Extreme and I'm ready to upgrade everything: 13900k, EVGA classified motherboard, EVGA PSU, EVGA CLX cooler and DDR5.

I will definitely get the 13900k, but I wish I could just wait for the 15900K.

 

[InVasMani]

My flight simulator is using an older Core i7 Extreme and I'm ready to upgrade everything: 13900k, EVGA classified motherboard, EVGA PSU, EVGA CLX cooler and DDR5.

I will definitely get the 13900k, but I wish I could just wait for the 15900K.

Be sure to check EVGA's website if it's the z670 Classified they have a bundle for it at $299 or did like a week ago with Z20 keyboard and a mouse I believe. The price alone is pretty good, but other inclusions make it a even better deal. NM seems that that ship sailed, but for $499 there is a bundle on the kingpin. Meanwhile if you get the kingpin on EVGA's website w/o the bundle not including the mouse and keyboard it's only $300's more. I guess Jensen was right the more you buy the more you do save!

 

[Why_Me]

My flight simulator is using an older Core i7 Extreme and I'm ready to upgrade everything: 13900k, EVGA classified motherboard, EVGA PSU, EVGA CLX cooler and DDR5.

I will definitely get the 13900k, but I wish I could just wait for the 15900K.

I'd consider the i9 13900 / 13900F due for release this January.

 

[Mussels]

@ModEl4
A few possible reasons:

1. Not all CPUs are equal and some may even run at significantly higher or lower voltages than the average.
2. As I mentioned earlier, when the CPU is running in a power-limited fashion, it is important that the DC Loadline be correctly configured, or results may be unusually higher or lower than they should be for the same reported Package Power, but almost nobody checks this out.
3. The AC Loadline used also influences the results. This is a motherboard setting for regulating load voltage (with Adaptive voltages) and help keeping the CPU at its built-in voltage-frequency curve. MSI notably uses a very high AC Loadline by default; other manufacturers (e.g. ASUS) tend to use a lower one.
4. Incompetence/sloppiness (wrong bios settings or testing methodologies).

EDIT: on a related note, HardwareUnboxed had testing issues:

Hardware Unboxed (@HardwareUnboxed)

Please note I will trim (remove) the power scaling section from our review. The data is accurate for the MSI motherboard used. BUT It appears as though MSI is very aggressive on voltage and doesn’t correct this when power limiting the CPU.

nitter.net

View attachment 266469

This definitely makes sense as MSI have been caught cheating this sort of thing several times now - Hell, HWinfo had a metric added just to catch them out for dishonest power reporting

They may blame intel XTU, but if that's the case it's even worse as any reviews using that software might also be reporting dishonest values

I'm on 8700K with 3080 and haven't noticed any "choking"

From what i've seen very few games max out the CPU cores, but i see a lot of reports from players on facebook about 100% usage on their 4 core/8 thread CPU's with a few modern FPS titles (call of duty/battlefield)

I don't play those so i havent experienced it firsthand

I'm researching this .. something strange is going on .. going away for the weekend in an hour though, more testing on monday

Might be related to the power values being limited/not limited as expected?

We've had dishonest motherboards in the past, and now evidence intels XTU software isn't working as expected