Overclocking on a i7-4790S?

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Why is it getting beaten by an FX-8350 in some of the benches he shows?


I never saw confirmation that all 4 cores were at 4.0 under stress?
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
I'll see if my E3 1231 V3 will lock at it's turbo speed this evening.

I have a z97-a board.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Yes, it looks for all the world like my Xeon is locked at 3.8, but further examination with XTU and Throttle Stop reveals TDP throttling going on.

So yes, you can lock the chip at it's turbo speed, but it is likely just TDP throttling.

No thermal throttling at all, though.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Okay, with further tweaking I got it to stay at 3.6 under stress with no TDP throttling.

So, I'm going to say that this is possible.

If you can avoid the TDP throttling, it will run all 4 cores at the higher multi.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Nope, I was wrong.

It just took longer to TDP throttle. It's still doing it.

It's also slightly slower than my earlier CPU-Z benchmarks with the chip.

EDIT: To clarify, these first attempts were with a B85 board.
 
Last edited:

iSkylaker

Member
May 9, 2015
143
0
76
I will fool around more with it later with the Z97 board.
Thank you for posting your finding, LT. I guess, due TDP throttling basically he got an slower/more unstable chip then, that's probably why the FX-8350 was scoring better.
I wonder if that "Long Duration Package Power Limit" he mentioned around 4:25 has something to do with the TDP limit, I have no clue but maybe adjusting it would probably improve stability and reduce throttling? although I suppose that TDP limit is hardcoded in the chip rather than being adjustable from the bios?
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
The ASUS Z97-A board has no problems. It runs the Xeon at 3.8 under full load. Just went to manual, sync all cores, set to 38. Reboot. And Bob's your uncle.

So far, so good.

Power set to balanced in Win10. Idles nice and cool. All four cores go to 3.8 when called.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Not seeing any TDP throttling with the Z97-A board and a Hyper 212Evo.
Runs like a champ! *

CPU-Z Stress:
y1c9t8j19jmf4wv6g.jpg

Burn Test:
o3gtne7lgxkuu3z6g.jpg

Passed Burn Test:
3ew5fysjy41gmms6g.jpg


* After it took me 15 minutes to remember that the Xeon had no IGP...
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
When I put my 4790K back in the Z97-A, it gave me the "dead CPU" LED on the board, and wouldn't do anything.

Moved the 4790K to another board. No go.

3 other Haswell chips work fine in both boards.

So apparently, my 4790K is now deceased, after less than one year of occasional use.
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
641
126
Don't touch BCLK on Sandy Bridge and newer.

Most chips can get from 0 to 5 out of bclck still stable. I have a i5 2500 (no k) running at 103 to get basically another 100mhz out of a locked chip. Its been running that way for over a year without a single problem
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
Most chips can get from 0 to 5 out of bclck still stable. I have a i5 2500 (no k) running at 103 to get basically another 100mhz out of a locked chip. Its been running that way for over a year without a single problem

You're of course aware that overclocking by BCLK overclocks everything? The PCIe controllers, SATA controllers etc.?

Running those out of spec is a bad idea, since it can cause silent data corruption and all sorts of other "fun" things.
 

DrMrLordX

Lifer
Apr 27, 2000
21,546
10,734
136
You're of course aware that overclocking by BCLK overclocks everything? The PCIe controllers, SATA controllers etc.?

Running those out of spec is a bad idea, since it can cause silent data corruption and all sorts of other "fun" things.

It could, but the odds of that actually happening are very low (perhaps non-existent). There's a reason why bclk bumps of 5 mhz are the most you can get out of Haswell system (barring a strap change on boards that support it). It'll go full-on unstable before anything "silently" corrupts itself.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
It could, but the odds of that actually happening are very low (perhaps non-existent). There's a reason why bclk bumps of 5 mhz are the most you can get out of Haswell system (barring a strap change on boards that support it). It'll go full-on unstable before anything "silently" corrupts itself.

It's still not something that should be generally recommended. Want to overclock? Just get a K. If you can already afford a 4790S, you can squeeze a 4790K in if you're going to OC, and get better stock performance too.

Besides, doesn't overclocking kind of defeat the purpose of the CPUs 65W TDP?
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
In my pics post, you can see that the Xeon barely went past it's TDP number while running IBT with all 4 cores at the 3.8Ghz turbo speed. Temps stayed in the mid 60's.

So, I don't think it was much of a strain on the chip.
 

DrMrLordX

Lifer
Apr 27, 2000
21,546
10,734
136
It's still not something that should be generally recommended.

Hey, whatever floats your boat. I don't fall on the conservative side when it comes to OC recommendations. I'd hesitate to do it on a file server with multiple HDDs/SSDs, but for a home machine, it's actually quite safe and stable. 105 mhz bclk is really nothing. USB devices can be pushed wayyyyyyy out of spec and work okay, and in my experience, PCIe cards behave very well up until a bclk of 119-120 mhz.

SATA is a slightly different issue, though I have had no stability/corruption issues using HDDs on standard SATA connectors at a bclk up to 110 or so mhz. Granted, that's on FM2+, but the intrinsic issues are the same.

Intel systems crap out at bclk speeds well below where you're going to start getting problems with subsystems tied to the bclk.

Want to overclock? Just get a K. If you can already afford a 4790S, you can squeeze a 4790K in if you're going to OC, and get better stock performance too.

Some people just have an S for whatever reason. So if they wanna know how to OC one, there's two ways to go about it, both of which work together:

Set the CPU to max turbo frequency, and
raise bclk to 105 mhz.

Besides, doesn't overclocking kind of defeat the purpose of the CPUs 65W TDP?

Maybe. That's up to the end-user to decide. More speed = more watts.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
Hey, whatever floats your boat. I don't fall on the conservative side when it comes to OC recommendations.

Neither do I. If you know what you're doing, and accept the risks, fine. But for someone new to overclocking, I still wouldn't recommend it. Not with completely multiplier unlocked Ks available at virtually the same price. Or 4790Ks which after all already do 4.0/4.4GHz at stock... :D
 

iSkylaker

Member
May 9, 2015
143
0
76
In my pics post, you can see that the Xeon barely went past it's TDP number while running IBT with all 4 cores at the 3.8Ghz turbo speed. Temps stayed in the mid 60's.

So, I don't think it was much of a strain on the chip.

hey LT thanks for letting me know, much appreciated.


@Thread, I was looking for performance numbers of those low-power chips when I found that video, I was just asking since it was always under my impression that you can't multiplier-overclock a locked chip, not even keep it at the turbo speed. I'm not really interested on that CPU.
 

Dufus

Senior member
Sep 20, 2010
675
119
101
3.8GHz should see over 200GFLOPS. Some locked Haswell's can be unlocked, others can use the top turbo bin on all cores when using microcode version 0x7 or less.
 

Zodiark1593

Platinum Member
Oct 21, 2012
2,230
4
81
3.8GHz should see over 200GFLOPS. Some locked Haswell's can be unlocked, others can use the top turbo bin on all cores when using microcode version 0x7 or less.

Is it possible to unlock some locked Haswells? How would that be done?