Intel Core i3-12100F

[Selaya]

you cant even overclock a 12600K on a H610, what on earth would make one believe you can OC anything at all w/ it?

 

[Assimilator]

you cant even overclock a 12600K on a H610, what on earth would make one believe you can OC anything at all w/ it?

Ask the other users in this thread...

 

[thelawnet]

I mean, the difference is going to be like this:

• Windows will take 7.8 seconds to boot instead of 7.6 seconds.
• Opening Chrome will take 0.82 seconds instead of 0.79 seconds.
• Loading a game level will take 12.4 seconds instead of 12.3 seconds.

The only instances where you'll see differences is when you copy several gigs of data from RAM to disk and back. Chances are you're not in the market for an ultra-budget platform if you're doing that regularly as the people who need a PCIe 4.0 SSD can usually justify the premium. Additionally, the cheapest PCIe 4.0 SSD on the market is still like a 25% more expensive than any of the obvious budget NVMe drive choices that you'd pair with a budget CPU and motherboard. The PCIe 3.0 SN570 performs very close to the PCIe 4.0 Atom50 but it's $30 cheaper. $30 is a lot in this segment.

Would it?

What I'm referring to is this:

"As of this writing, VERY few motherboards use the H670 client chipset, making the B660 Intel's mainstream desktop chipset for those with a "locked" 12th Gen Core processor ....
The H610 is the bare entry-level chipset. You lose out on memory overclocking, only get Gen3 PCIe connectivity across the board, and no CPU-attached NVMe."

from the review.

In fact memory overclocking doesn't seem to be a loss, and it seems misleading to say you get only Gen3 PCIe connectivity across the board.

You have in all cases a PCIE 5.0 or 4.0 x16 slot from the CPU, though the cheaper boards will run it at PCIE 4.0.

You have a 4.0 x4 slot from the CPU for NVMes. This is disabled on the H610.

This leaves the H610 with 8 x PCIE 3.0 lanes, which are spent:

* NVME 3.0 x4, NVME 3.0 x 2, PCIE 3.0 x 1 as on the Asus Prime E/ A (7 lanes, and USB, SATA, GBE use other flex io lanes)
* NVME 3x4, PCIE 3 x1, PCIE 3x1, as on the Asrock HDV/M.2 (6 lanes, etc.)

etc.

So realistically the fullest use of the H610 chipset would be something like the Asus Prime E or A, but with a second x1 PCIE slot.

So the H610 chipset only supports 20 out of 48 HSIO lanes, and it disables lane 17-20 from the CPU, which should be used for your first M.2 device.

Now the question that I'm asking is the review notes that there is no CPU-attached NVME and implies this is a bad thing.

So instead of plugging your budget PCIE 3.0 x4 NVME into a slot connected to the CPU, which supports 4.0 x4, but runs at 3.0 x4, because that's the speed of the device, you instead plug in the NVME to a slot connected to the chipset, also a 3.0 x4, where it connects to the CPU via the DMI lanes, which are 4x DMI 4.0, which as I understand it is equivalent to 4 x PCIE 4.0, or 8 x PCIE 3.0.

I suppose it depends what exactly you were doing, e.g., copy A->B would not touch the CPU but 'load a game into memory' might.

On a B660 board, typically with two M.2 slots at 4.0 x4, this means one is CPU-attached and one is chipset-attached. So that is basically exactly the same issue - does it in fact make any performance difference whether you attach an M.2 device directly to the CPU or to the chipset. If not, it would seem like the review should read "NVME speed is limited to PCIE 3.0", not "no CPU-attached NVME"

(note: the review chart has a few omissions IMO :

H610 is 1 DIMM per channel, otherwise 1 or 2
RAID should be noted: H610 - none, B660 - SATA, H670/Z690 - SATA/NVME
Max displays is 3 on H610, otherwise 4)

 

[Chrispy_]

Ask the other users in this thread...

I think you're confusing all the discussion about the value proposition of this i3 in terms of additional platform cost with the BCLK overclocking discussion in the other thread.

At present there are no DDR4 B660 or H610 boards with external clockgen for BLCK overclocking. That means that to BCLK overclock a $105 i3 you need - as the lowest-cost minimum requirements - a $250 ROG Strix G/F B660 board and a minimum of $170 of unobtainable DDR5. $420 of hard-to-get platform cost is irrelevant to a $105 CPU that's likely to be paired with a ~$90 board and $60 of DDR4-3200.

 

[Dyatlov A]

Very nice CPU, i will buy a few and test the best for myself. But why he is not writing if he overclocked with the DDR4 or DDR5 board?

 

[W1zzard]

But why he is not writing if he overclocked with the DDR4 or DDR5 board?

I overclocked using the DDR5 board, the DDR4 board has no external clock generator

 

[seth1911]

The best performace/€ CPU for Gaming is useless with no GPU.

So long im using a A10 6800K

 

[Selaya]

Would it?

What I'm referring to is this:

"As of this writing, VERY few motherboards use the H670 client chipset, making the B660 Intel's mainstream desktop chipset for those with a "locked" 12th Gen Core processor ....
The H610 is the bare entry-level chipset. You lose out on memory overclocking, only get Gen3 PCIe connectivity across the board, and no CPU-attached NVMe."

from the review.

In fact memory overclocking doesn't seem to be a loss, and it seems misleading to say you get only Gen3 PCIe connectivity across the board.

You have in all cases a PCIE 5.0 or 4.0 x16 slot from the CPU, though the cheaper boards will run it at PCIE 4.0.

You have a 4.0 x4 slot from the CPU for NVMes. This is disabled on the H610.

This leaves the H610 with 8 x PCIE 3.0 lanes, which are spent:

* NVME 3.0 x4, NVME 3.0 x 2, PCIE 3.0 x 1 as on the Asus Prime E/ A (7 lanes, and USB, SATA, GBE use other flex io lanes)
* NVME 3x4, PCIE 3 x1, PCIE 3x1, as on the Asrock HDV/M.2 (6 lanes, etc.)

etc.

So realistically the fullest use of the H610 chipset would be something like the Asus Prime E or A, but with a second x1 PCIE slot.

So the H610 chipset only supports 20 out of 48 HSIO lanes, and it disables lane 17-20 from the CPU, which should be used for your first M.2 device.

Now the question that I'm asking is the review notes that there is no CPU-attached NVME and implies this is a bad thing.

So instead of plugging your budget PCIE 3.0 x4 NVME into a slot connected to the CPU, which supports 4.0 x4, but runs at 3.0 x4, because that's the speed of the device, you instead plug in the NVME to a slot connected to the chipset, also a 3.0 x4, where it connects to the CPU via the DMI lanes, which are 4x DMI 4.0, which as I understand it is equivalent to 4 x PCIE 4.0, or 8 x PCIE 3.0.

I suppose it depends what exactly you were doing, e.g., copy A->B would not touch the CPU but 'load a game into memory' might.

On a B660 board, typically with two M.2 slots at 4.0 x4, this means one is CPU-attached and one is chipset-attached. So that is basically exactly the same issue - does it in fact make any performance difference whether you attach an M.2 device directly to the CPU or to the chipset. If not, it would seem like the review should read "NVME speed is limited to PCIE 3.0", not "no CPU-attached NVME"

(note: the review chart has a few omissions IMO :

H610 is 1 DIMM per channel, otherwise 1 or 2
RAID should be noted: H610 - none, B660 - SATA, H670/Z690 - SATA/NVME
Max displays is 3 on H610, otherwise 4)

yeah but this is a product for the every single fucking cent counts bracket of the stack and given that even on h610 you can do like, 3200-c16 these are all fucking luxury features. i'd stick in a SATA ssd over an nvme if it means knocking another $5 off the total build's price.
so pairing it w/ the most bottom-of-the-barrel h610 makes perfect sense for something like an 12100f.

 

[tadoritz]

Power draw turned out to be similar to 12400@5.2GHz tested by der8auer and Hardwareluxx. Is it OK?

Reviewer	CPU	Power Draw
Hardwareluxx	12400@5.2GHz	131W on CPU in R23
der8auer	12400@5.2GHz	138W on CPU in R20
TechPowerUp	12100F@5.2GHz	174W Total System in R23

Voltage is about the same.

In comparison with 5600X, 3900X and 10900K:
Hardwareluxx: 12400@5.2GHz = 131W = 5600X+55W = 3900X-10W = 10900K-85W (on CPU)
TechPowerUp: 12100F@5.2GHz = 174W = 5600X+48W = 3900X-27W = 10900K-72W (Total System)

 

[bug]

Power draw turned out to be similar to 12400@5.2GHz tested by der8auer and Hardwareluxx. Is it OK?

Reviewer	CPU	Power Draw
Hardwareluxx	12400@5.2GHz	131W on CPU in R23
der8auer	12400@5.2GHz	138W on CPU in R20
TechPowerUp	12100F@5.2GHz	174W Total System in R23

Voltage is about the same.

In comparison with 5600X, 3900X and 10900K:
Hardwareluxx: 12400@5.2GHz = 131W = 5600X+55W = 3900X-10W = 10900K-85W (on CPU)
TechPowerUp: 12100F@5.2GHz = 174W = 5600X+48W = 3900X-27W = 10900K-72W (Total System)

It's ok, given how CPUs work these days. But it's not optimal.
The 12400F has two more physical cores, it should draw more power. If it doesn't, it means its cores run slower than those of the 12100F (probably thermal throttling).

 

[tadoritz]

It's ok, given how CPUs work these days. But it's not optimal.
The 12400F has two more physical cores, it should draw more power. If it doesn't, it means its cores run slower than those of the 12100F (probably thermal throttling).

Wouldn't it be unprofessional to throttle CPU so hard, and are there any signes of throttling? The scaling of Cinebench results compared to stock is also about the same in all three reviews.

 

[bug]

Wouldn't it be unprofessional to throttle CPU so hard, and are there any signes of throttling? The scaling of Cinebench results compared to stock is also about the same in all three reviews.

It's not unprofessional, it's how CPUs work: when they get too hot, they dial it back.
As for Cinebench scaling, I wouldn't know, I have run that thing.

 

[thelawnet]

yeah but this is a product for the every single fucking cent counts bracket of the stack and given that even on h610 you can do like, 3200-c16 these are all fucking luxury features. i'd stick in a SATA ssd over an nvme if it means knocking another $5 off the total build's price.
so pairing it w/ the most bottom-of-the-barrel h610 makes perfect sense for something like an 12100f.

Ok, but that wasn't the point I was discussing.

Specifically it was suggested in the review that you should pair this thing with a B660, not a H610, on the basis that having your SSD, likely SATA or at best PCIE 3.0 x4, connected to the PCH, rather than to the CPU, is a significant issue.

I presume it's not and therefore H610 is absolutely fine, in that it's not really reasonable to expect deluxe features for every possible future peripheral when you bought the poverty CPU, so if you start off with say an M.2, a 12100f, and then in the future need more space, well go buy a Crucial MX500 SATA SSD.

I mean, I'm not really sure what the +- is on buying a 12100 vs a 12100f, as opposed to an H610 vs a B660. Like if you could pick one... (Obviously contingent on also having a GPU). One big advantage of the 12100, is that if you're building a PC bit by bit, then you can get started with a 12100 and a H610 and then add a more suitable GPU later on. I mean sure, the Intel HD 730 IGP( or whatever it's called), is that while it will suck for Cyberpunk 2077, but, like it's a whole lot better than no PC, and you can play older games, indie games, etc., no problems, and then when you have more cash you can add a 6500 XT or something.

 

[Selaya]

[ ... ] on the basis that having your SSD, likely SATA or at best PCIE 3.0 x4, connected to the PCH, rather than to the CPU, is a significant issue.
[ ... ]

yeah. no.
until we get directstorage (coming in 2031) it's literally a nonissue. a luxury problem of a luxury problem. you're far better off spending your money on your gpu.

my priority for such a build would be like:
decent psu > gpu > gpu qol > cpu > board qol
as long as you've got an ssd (pref not a qvo, but even that'll do in a pinch) at all you're set basically; the ssd's performance impact's measured in like the 0.x%s

 

[tadoritz]

It's not unprofessional

It will be, if you test CPU with underperforming cooler or mobo and keep quiet about it.

Hardwareluxx assure that they didn't reach throttling temp.

We cooled the Core i5-12400 during base clock overclocking using a Noctua NH-U12A. However, this still managed to keep the processor below the threshold of thermal throttling. With 97 °C under constant load, but only very slightly. - Google Translate

 

[Assimilator]

yeah. no.
until we get directstorage (coming in 2031) it's literally a nonissue. a luxury problem of a luxury problem. you're far better off spending your money on your gpu.

my priority for such a build would be like:
decent psu > gpu > gpu qol > cpu > board qol
as long as you've got an ssd (pref not a qvo, but even that'll do in a pinch) at all you're set basically; the ssd's performance impact's measured in like the 0.x%s

A QVO will be a massive and obvious improvement over a HDD. Hell, even a non-name-brand using rejected NAND chips will be a massive and obvious improvement!

It will be, if you test CPU with underperforming cooler or mobo and keep quiet about it.

Hardwareluxx assure that they didn't reach throttling temp.

Did you read the "Test Setup" page of the review?
Did you read the "Temperatures" page of the review?

 

[tadoritz]

Did you read the "Test Setup" page of the review?
Did you read the "Temperatures" page of the review?

We were talking about 12400s. Bug suggested that Hardwareluxx's and der8auer's 12400s were throttled, not this 12100.

The 12400F has two more physical cores, it should draw more power. If it doesn't, it means its cores run slower than those of the 12100F (probably thermal throttling).

 

[lexluthermiester]

Well folks, we have the perfect budget king CPU, never thought I'd say congrats to Intel, great review Wizz.

View attachment 233922

I have to agree. This is damn good for a 4c/8t i3.

 

[Cowboystrekk]

This is disapointing! SA-voltage is stuck at 0.9-0.96V if you run at locked CPU. This will allow for 3400-3733 OC in gear 1, in many cases you can`t even get 3600 stable. Shame on you Intel.

From MSI internal test, so far the result is as below,
*All result is done by MP CPU, QS CPU might have inconsistent result as MP CPU*

• Non-K + B chipset = cannot increase VCCSA (it sits around 0.91x to 0.92xV depending on IMC quality)
• Non-K + Z chipset = cannot increase VCCSA (it sits around 0.91x to 0.92xV depending on IMC quality)
• K + B chipset = can increase VCCSA (This scenario was not confirmed by Intel, but so far it worked)
• K + Z chipset = can increase VCCSA

MSI's Latest Z690 & B660 Motherboard BIOS Improves Intel Alder Lake Non-K Memory Compatibility (wccftech.com)

 

[Lionheart]

I have to agree. This is damn good for a 4c/8t i3.

Yeah for sure, problem is what GPU could you pair it with without ripping yourself off? Seems like the used market is the only way to go cause the 6500XT is garbage, GTX 3050 are a tad better but still OP

 

[lexluthermiester]

Yeah for sure, problem is what GPU could you pair it with without ripping yourself off?

I would say up to the 3060ti before any real CPU bottlenecking starts to show.

A QVO will be a massive and obvious improvement over a HDD. Hell, even a non-name-brand using rejected NAND chips will be a massive and obvious improvement!

QLC durability is garbage compared to ANY HDD. Speed increases be damned.

 

[tadoritz]

Update.

Reviewer	CPU	Power Draw
Hardwareluxx	12400@5.2GHz	131W on CPU in R23
der8auer	12400@5.2GHz	138W on CPU in R20
TechPowerUp	12100F@5.2GHz	174W Total System in R23
New data...​	...​	...​
Hardwareluxx	12100F@5.2GHz	92W on CPU in R23
Lawrence Timme	12100F@5.2GHz	96W on CPU in R15
JAWARA media	12100F@4.6GHz	73W on CPU in R23

Hardwareluxx 12100F - 8935 pts in R23 (7508 in stock per reviewer)
JAWARA media 12100F - 9499 pts in R23 (8490 in stock per reviewer)
[TechPowerUp 12100F - 10748 pts in R23 (8560 in stock per reviewer)]

Lawrence Timme 12100F - 1534 pts in R15 (1346 with BCLK=110 MHz per reviewer)
Hardwareluxx 12100F - 1344 pts in R15 (1050 in stock per reviewer)

JAWARA media 12100F - 3641 pts in R20 (3249 in stock per reviewer)
Hardwareluxx 12100F - 3623 pts in R20 (2893 in stock per reviewer)

 

[Rockfella.killswitch]

I still don't see the need to upgrade lol.

 

[Cowboystrekk]

Update.

Reviewer	CPU	Power Draw
Hardwareluxx	12400@5.2GHz	131W on CPU in R23
der8auer	12400@5.2GHz	138W on CPU in R20
TechPowerUp	12100F@5.2GHz	174W Total System in R23
New data...​	...​	...​
Hardwareluxx	12100F@5.2GHz	92W on CPU in R23
Lawrence Timme	12100F@5.2GHz	96W on CPU in R15
JAWARA media	12100F@4.6GHz	73W on CPU in R23

Hardwareluxx 12100F - 8935 pts in R23 (7508 in stock per reviewer)
JAWARA media 12100F - 9499 pts in R23 (8490 in stock per reviewer)
[TechPowerUp 12100F - 10748 pts in R23 (8560 in stock per reviewer)]

Lawrence Timme 12100F - 1534 pts in R15 (1346 with BCLK=110 MHz per reviewer)
Hardwareluxx 12100F - 1344 pts in R15 (1050 in stock per reviewer)

JAWARA media 12100F - 3641 pts in R20 (3249 in stock per reviewer)
Hardwareluxx 12100F - 3623 pts in R20 (2893 in stock per reviewer)

Different voltages and powerlimits accounts for much of the difference we see I bet. Stock 12400 on hardwareluxx scores 11100 on R23 using 72W. On my setup I get 11700 stock (80W short, 65W long, last 10sec of test switches to long), but using Asus pwr unlock setting I get 12400 at 75W. Using Asus settings voltage stays at 1000mv avg in CB23, using stock it stays at 1050mv avg when at 80W short and 960mv at long.

 

[Cutechri]

Budget king right here. AMD where's your budget products now?