Intel Core i5-13400F

[Wasteland]

The stability test is over 30 days since I use it. I don't allow system crashes in online games. A good word can put 12500, but it can't. It now runs with other memories at 3200MHz because that's what xmp has.
Unsafe is that situation where you have to increase the voltage above the specifications to maintain stability at a certain frequency.

I don't want to prove anything. Apart from HU, all reputable reviewers have tested Alder and Raptor at 3600 without problems. Keep clinging to this nonsense with "unsafe".

Ok, link me to an article from a "reputable reviewer" who tests locked Alder/Raptor Lake CPUs with DDR4-3600 in Gear 1.

And while you're at it, give me a direct quote where I used the word "unsafe." You can't because I didn't use that word. What I actually said is that it's irresponsible to tell people who may not know about memory overclocking that they can run at unstable speeds. The issue there isn't hardware "safety." The issue is inconvenience and data corruption. possibly a little wasted money.

 

[Gica]

One is here. If you want more, we go further because many have tested DDR4 vs DDR5, especially with Alder.

"Safe" I said, and you considered it irresponsible.
If so, all those who use memories clocked over DDR4 3200/DDR5 4800 (Alder), or 3200 DDR4/DDR5 5600 (Raptor i7, i9 or 13600K), or 5200 DDR5 for Ryzen 4, are irresponsible. In your opinion, of course. Even @W1zzard is irresponsible because he did not test the processors according to the manufacturers' specifications.

 

[Wasteland]

Incidentally, this Techpowerup review that Gica quoted earlier was written based on testing with an i9-12900k, which is not a locked CPU. Not only that, it was written before any of the locked Alder Lake SKUs launched. These facts point to why I harp on this issue. Most people who google for info on what memory to buy with their locked CPU will naturally encounter a lot of talk about unlocked-CPU configurations, because reviewers tend to concentrate on those models. Info on the memory-speed limitation on locked Intel CPUs is hard to come by unless you're specifically looking for it.

It caused me some headaches with my own CPUs. This isn't an AMD vs Intel issue. Good grief. This is genuinely just an issue of recommending to people the proper RAM for their build, and by proper I mean the RAM config that will run as fast as possible without causing problems.

I'm not blaming W1zzard for recommending DDR4-3600 for Alder Lake; how could he have done otherwise in November of 2021? But I doubt very much that he would agree with Gica's position that the VCCSA limitation on locked LGA 1700 CPUs is just some baseless pro-AMD-fanboy conspiracy theory put forward by Hardware Unboxed, Buildzoid, and Micro-Star International. Seriously, this convo is like taking crazy pills.

One is here. If you want more, we go further because many have tested DDR4 vs DDR5, especially with Alder.

View attachment 289143


"Safe" I said, and you considered it irresponsible.
If so, all those who use memories clocked over DDR4 3200/DDR5 4800 (Alder), or 3200 DDR4/DDR5 5600 (Raptor i7, i9 or 13600K), or 5200 DDR5 for Ryzen 4, are irresponsible. In your opinion, of course. Even @W1zzard is irresponsible because he did not test the processors according to the manufacturers' specifications.

See above. That is not a review of a locked CPU, and in fact it predates the launch of locked Alder Lake CPUs. Keep trying.

You clearly have a language barrier issue if you think I argued that you're irresponsible for running your memory at a given speed. I feel like I've said this a dozen times, but once more for old time's sake: I don't care what you do with your own computer. Nobody cares. Telling other people that they should buy memory and run it at likely unstable speeds, and all because you're fixated on some ludicrous fanboy war--that is what is irresponsible.

 

[Gica]

12400 DDR4 3800MHz Gear 1 is it good for you? And these are the first samples. Today's Alder and Raptor architecture is much more refined.

If it has a problem with running at 3600, it can quickly switch to lower speeds. There are no big price differences between 3200 and 3600 and with the 3600 you can run at 3200 with lower latencies.
The data coming from the media networks show that there are rather problems with the memory and/or the motherboard, not at all with the processor. There are millions of system configurations and sometimes incompatibilities appear.
RAM X + MoBo Y = love... or not

I can go on as long as you want, but I'm coming down to your level.
The discussion is closed.

 

[Wasteland]

Excellent, you found one. So between Tomshardware's one chip (they bought a single 12400 before retail launch for that review--they couldn't even supply 100% accurate description of the specs because the specs hadn't been announced yet), and your chip (assuming it is stable), we have two whole examples! That definitely proves that HUB, MSI, and Buildzoid are just making stuff up. The posters here, including me, must've imagined their blue screens too.

 

[Gica]

The HU administrators made their accounts on TPU, LOL.

 

[Gica]

Session: 5+ hours (gaming, installations, www, etc.)
Y-Cruncher all test: 1h (in parallel with installing a game)
Undervolt 0.2V DDR memory and CPU memory controller (VDDQ TX)
Timmings: 18, 22, 22, 42 -> 17, 20, 20, 38
I can't keep the computer running for a month to prove to you that you are talking nonsense. I repeat: 12500 also behaves identically. I have two processors, both work perfectly at 3600MHz DDR4. Am I the cloudiest person on the planet or the "unlucky" ones, a small minority, have other problems and not the VCCSA blocking?
Search the net for testing the 500 Raptor processors. Even the weakest can run stably at 7600 (35.7%), and they are a small minority, below 5%. Most can carry 8000+, that is 43% over specifications.
For DDR4, the jump from 3200 to 3600 means only 12.5%.

 

[NRANM]

Finally! The review has been updated with corrected numbers for the 7600 and 7600X. I've also retested the other Zen 4 CPUs on the 9922 BIOS with the newest chipset drivers.

While Ryzen 7600/7600X have gained quite a bit in gaming performance and power, the outcome of the conclusion remains unchanged

Sorry to re-open the topic, but were the results for the VirtualBox virtualization tests also wrong? If so, were they corrected? The Ryzen 7600's results from its own/original review seem quite different from the results in this review.

 

[rv8000]

@W1zzard

Is the efficiency chart at 1080p correct? You list an average of 217fps for the 7600 at an average of 49w, which would be 4.43 frames per watt. Then you list the 13400f at 185fps at an average of 43w, which would be 4.3 frames per watt.

 

[gasolin]

I'm not sure if it's actually that close. The 13400F comes with a stock cooler, whereas the 12600K does not. Getting a decent cooler for the 12600K will widen the price gap.

NO thermalright peerless assasin 120

According to toms harware it's the best gamer cpu and it doesn't use much power

Here are the Best CPUs for Gaming that you should buy in 2025

Here is the best CPU for gaming.

www.tomshardware.com

Good review. What grinds my gears in not quoting max all core turbo frequency anywhere. I fell into this trap with a 12400 I bought lately, it's sold as a 4.4GHz CPU but good luck seeing 4.4 GHz unless you start a single-threaded load and limit affinity to a single core. I've never seen it go beyond 4.2GHz for a split second and for any realistic load it's a 4GHz CPU. Even Intel seems to be ashamed of that, hiding actual frequencies from potential customers with the usual "up to" rubbish.
Otherwise a good CPU, just scammy marketing.

Try increasing power limit

 

[gasolin]

why is my score so low ? Why is multiplier ony x41 instead of 45 ?

 

[Chrispy_]

why is my score so low ? Why i multiplier ony x41 instead of 45 ?

There is no "score". What on earth are you talking about?! The multiplier varies from moment to moment. If you go to the AIDA CPU page you can see it fluctuate in real time as background services use the CPU in varying amounts.

Your read/wite/copy/latency results are different, but that's because you're using different memory at different timings, so of course it's going to be different. You're also using a different motherboard and the exact same kit of RAM and the exact same CPU can provide different results in different motherboards. They'll sometimes even differ in the same motherboard on different BIOS versions.

Don't overthink it, and if you want to tune your DDR5 make a full backup of your OS or be prepared to have it corrupted by unstable memory timing as you learn how to tune DDR5. Nobody gets it perfectly right from just reading an article, and nobody learns the stability limits without reaching instability, at which point something as simple as a windows update with unstable RAM timings can screw up your OS.

 

[gasolin]

6000mhz cl36 vs 6400mhz cl32 i should score higher

 

[THU31]

Your write/copy bandwidth is higher than the TPU result, because of the higher memory clock, but your latency is also higher, probably because of timings. Your CL timing is lower, but all the other timings are higher.

Your L3 cache performance seems very low (mine is similar to the TPU result), which is weird, considering your north bridge clock is higher. I don't remember if there are other settings affecting L3 cache performance other than the clock speed. You might want to look into that.

And the 13400 has a max boost clock of 4.1 GHz for all cores. 4.5 GHz is only for one or two cores under a low load (but you need all power-saving features enabled to achieve that, including core parking as far as I know, which is not recommended on desktop).

 

[Chrispy_]

6000mhz cl36 vs 6400mhz cl32 i should score higher

Those numbers are only half the timings.

6000 36-36-36-76
vs
6400 32-39-39-102

The 6000 kit is lower latency than your 6400 kit overall. CL32 applies to access (read) speeds only but it's RAM, not ROM, so it needs to more than just read.

Stop worrying about it, more bandwidth often comes at the cost of looser timings, unless you want to tune the RAM manually or spend stupid money on a kit with very tight timings embedded into the XMP profile out of the box.

Bandwidth is more important than latency for Intel platforms anyway - If you were on AMD I'd tell you to drop your clocks to 6200MT/s and tighten up those timings, but you're not, so don't worry about the higher latency and remember what I said last time:

Don't overthink it

 

[gasolin]

bus speed 99.76 not much diffrence (4090mhz) but only runs 4090 (asus mb) gigabyte runs 100

So 6000mhz 30-36-36-76 is better than 6400 32-39-39-102 ?

 

[Chrispy_]

bus speed 99.76 not much diffrence (4090mhz) but only runs 4090 (asus mb) gigabyte runs 100

So 6000mhz 30-36-36-76 is better than 6400 32-39-39-102 ?

"Better" is a very complicated definition in the context of RAM. Better in some applications, worse in others. Better with some CPUs, worse with others.

What is better - a Ferrari or a pickup truck? Because the answer depends on whether you at at a race track or or trying to move one tonne of cement across town? The answer depends on the situation. Stop thinking of it in terms of black and white, all the answers are shades of grey.

The last of the numbers is TRAS, and your 102 TRAS is very slow compared to the 76 TRAS of the test kit. On the other hand your kit is running faster and has more bandwidth. As a rough rule of thumb, Intel 13th Gen prefers bandwidth to low latency, so the 6400 vs 6000 is the most important number for you. Sure, a lower latency kit would be an improvement, but not by a large margin.

Your L3 cache performance seems very low (mine is similar to the TPU result), which is weird, considering your north bridge clock is higher. I don't remember if there are other settings affecting L3 cache performance other than the clock speed. You might want to look into that.

Nah, I suspect that's just AIDA64 being crap.

L3 cache results are incredibly sensitive to power states during the moment of test. Run the test multiple times and on systems set to balanced power profiles with idle parts of the CPU being dropped to lower power states and you'll get a different number for L3 cache speeds each time.

AIDA64 seems to be less useful and less relevant with each passing generation. IMO it belongs to the same era as it's UI, IE the Windows 95-Windows XP era.

 

[gasolin]

I also see on the qvl list a 13 gen k,kf can handle mucher higher ram speed 7000 mhz and higher, my non k cpu (basically a non overclockable 12600k) is limited to 6400mhz

General a 13 gen k cpu offers more ram,memory flexability if you wanna go up to the limit of your mb's ram speed, 13 gen k and non k cpus don't have the same ram speed limit

about the best i can get

 

[JustBenching]

I also see on the qvl list a 13 gen k,kf can handle mucher higher ram speed 7000 mhz and higher, my non k cpu (basically a non overclockable 12600k) is limited to 6400mhz

General a 13 gen k cpu offers more ram,memory flexability if you wanna go up to the limit of your mb's ram speed, 13 gen k and non k cpus don't have the same ram speed limit

about the best i can get

View attachment 401926

You can certainly drop the latency by a lot. Tighten down your TRFC and max out your Trefi (65535). I've got 47ns latency on a 12900k

 

[gasolin]

G.Skill Ripjaws S5 DDR5-6400 - 32GB - CL32

F5-6400J3239G16GX2-RS5K

I guees it depends om the speed of the cpu and that it has to be quad channel memory

I will never come colse to 50ms in latency with my 13400f and asus Z790 mb

can a 13400 handle 4x16 6400mhz cl 32 or 6600mhz cl 32 ? (6400mhz cl 30 and 6800 mhz cl 32 isn't stable)