Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Unintended consequences of Intels 45nm process
- - By Nick (*****) Date 2007-12-20 11:57
Is this the beginning of the end for us overclockers?

If anyone has a QX9650 can they confirm this story?

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3184&p=4

If it's true is it just a consequence of reducing size or will AMD be exempt?  :)

Best regards,
Parent - - By Banned for Life (Gold) Date 2007-12-20 18:43
It's not obvious that this has anything to do with the 45 nm process. It sounds more like Intel implementing a more sophisticated protection system that looks at the cores individually and reduces core frequency and voltage individually in a manner not subject to user control.

Its possible that Intel did a cost/benefit  analysis of the enthusiast market and decided that allowing overclocking doesn't justify the increased number of warranty returns. This would be less surprising if it happened with a lower end processor and it may have been unintentional with the QX9650.

Alan
Parent - - By NATIONAL12 (Gold) Date 2007-12-20 22:47
i know kullberg has bought these chips.they cost a lot of money.i will be very interested on his comments on overclocking them.
Parent - - By Lukas Cimiotti (Bronze) Date 2007-12-20 22:54
I ordered one, but i didn't get any so far :(
Parent - - By NATIONAL12 (Gold) Date 2007-12-20 23:10
let us know if you eventually get them. but this site that bigmomma posted is well respected and a small worry.speed merchants for intel represent .00001 of intels turnover and publicity. perhaps some of you guys like alan and yourself could comment on their posts in their websites and give direct feedback to the rest of the computer world.
Parent - - By Lukas Cimiotti (Bronze) Date 2007-12-31 16:02
I've got it now, but I'm not very happy with it. On my Asus P5K deluxe it only reaches 4 GHz with ~1.3 V VCore. Raising VCore doesn't help a bit and improoving cooling doesn't help either. I'm using a chiller cooling with 15°C water temp. So I am quite disappointed about that CPU. One problem might be the mainboard as there is a huge drop of voltage on load. I'm thinking about getting a different one, but I haven't decided that yet.
Maybe I'll wait for Intel Skulltrail.

Regards,
Lukas
Parent - By Nick (*****) Date 2007-12-31 16:50

> One problem might be the mainboard as there is a huge drop of voltage on load


That mainboard has a CPU voltage dampener setting available in BIOS which when enabled is meant to greatly reduce vdroop.   You could try the pencil mod also.
Parent - By NATIONAL12 (Gold) Date 2007-12-31 18:44
very disappointing for you.i am not literate enough about computers to be able to offer any advice,please let know if you if you manage to improve performance. these chips are not cheap. regards PAUL.
Parent - - By Phil Harris (***) Date 2007-12-31 21:00
I would be very surprised if the BIOS was anywhere near optimised for the QX9650 yet, so I expect there is plenty more to come from that combination, especially with the voltage damper included. From what I have read, the damper has yet to be really effective, but later BIOS revisions should improve that.

My own experience of Asus boards is that vdroop can be very pronounced at high overclocks. On my old P5E I found the pencil modification made a huge difference to it's stability, so a little research in that area will help I'm sure.

Phil
Parent - - By Banned for Life (Gold) Date 2007-12-31 22:23
I have been using Gigabytes boards rather than Asus and they have a dynamic system to address the Vcore droop under load that seems to work fairly well. Are the ASUS boards deficient in this area?

Regards,
Alan
Parent - By Lukas Cimiotti (Bronze) Date 2007-12-31 23:54
I think, you are right. I've got an old Gigabyte GA-965P-DQ6. It was best for overclocking my QX6700. Maybe I'll order the Gigabyte GA-X38-DQ6. But I only purchased that QX9650 for testing purposes, so I'll won't spend too much effort to overclock it to the absolute maximum.

Regards,
Lukas
Parent - - By RFK (Gold) Date 2008-01-02 22:48
Is there any chance that other hardware element-such as but not limited to -audio/video component- could in some why be affected by the 45nm technology that would have to go through a natural transition period.
Parent - By Banned for Life (Gold) Date 2008-01-02 23:08
Audio is decoupled from the processor so it rarely causes any problem (I turn it off in the BIOS on my chess computer anyway). Video can be an issue if you are trying push things for gaming, but this doesn't apply to chess. Instability when trying to OC a chess computer is almost always associated with one of the following:

- Excess heat in processor, memory, northbridge chipset, or on the motherboard itself,
- Insufficient voltage for the processor or memory or northbridge chipset (or unstable voltage), or
- Excess frequencies on the processor, FSB, or memory, or overly tight memory timing.

Regards,
Alan
Parent - - By Lukas Cimiotti (Bronze) Date 2007-12-21 09:47
I had that same problem with my 2 Xeon X5365 (taken from a Mac Pro - TDP 150W) on an Intel S5000PSL using Verax X21 coolers. CPU usage showed always 100% but temperatures of cores were jumping up and down (in a safe range - no cpu thermal throtteling). Overall performance was not good.
That problem was in fact caused by overheating of voltage regulators as those Verax coolers have a horizontal air flow. I now use Intel boxed coolers - temperatures of the CPUs are not better, but voltage regulators are cooled much better.

In the case of that QX9650 things might be different.
But i think, Anandtech's approach to solve that problem wasn't optimal. I think, better cooling of VRMs or trying a different mainboard would have been better.

Regards,
Lukas
Parent - - By M ANSARI (*****) Date 2007-12-22 09:36
Interesting article ... I really don't like the fact that there is no warning or no data that the processors are being throttled down.  This can cause havoc if you are trying to tune the system by simply looking for 100% CPU work load and frequency speed as the final data.  I guess this is why to really check performance of a system you have to throw countless different benchmarks at it to really check the performance.  I have sometimes felt that my highly overclocked computer sometimes plays inconsistently than normal ... while the less overclocked one seems more predictable.  Maybe there is something there.  Need to check it on benchmarks that use all cores at 100% for long periods of time (say 6 hours) ... as it is benchmarks usually run for short periods or simply check for stability at longer periods.  Does anyone know of a long timed performance benchmark?
Parent - - By Banned for Life (Gold) Date 2007-12-22 16:50
Probably the best way to see if your processors are throttling down is to do infinite analysis of a position with no table base accesses and check later to ensure the 4 core temps are stable after they reach equilibrium. Any internal throttling, frequency alone or frequency and voltage, will certainly show up in this type of test.

Alan
Parent - - By Nick (*****) Date 2007-12-23 12:07 Edited 2007-12-23 12:20
I was in London yesterday (XMAS shopping) and Santa gave me a deal on one of these QX9650's plus an ASUS Maximus Formula, Tottenham Ct Rd.   I stuck the board and chip into my water rig, did the vdroop pencil mod, and upped the FSB (400), vcore (1.58v in bios) and mult (x11) so that 4.4 ghz was stable and did the test Alan suggested.  I didn't notice any change in core temps (using coretemp), vcore or frequency (using cpu-z) once they had reached a steady state of 62,51,56,55, I gave up the test after four hours.   Hardly scientific I know but I'm not (obviously) seeing the effect mentioned in the article - at 4.4ghz anyway. 
Parent - By Banned for Life (Gold) Date 2007-12-23 17:25
Wow, sounds like you had a good (and busy) day yesterday.  Your testing actually sounds pretty scientific to me and I suspect that the article either had some other non-chip issue, or that the problem only occurs when core temp is above 62C.

Regards,
Alan
Parent - By BB (****) Date 2007-12-24 01:24
Santa gave me a deal

You mean he's stopped giving away gifts and entered the market economy now? :-P
Parent - - By RFK (Gold) Date 2007-12-23 18:04 Edited 2007-12-23 18:12
Interestingly enough, I am seeing a gradual decrease in price in all chips over the last few months. I am sure for guys with the big bucks like you this means absolutely nothing. But, for us guys on the poverty line it is a welcome sign. It also flags a possible market signal. Unless of course you have already on order the 45nm Hi-k Intel® Core™2 Extreme quad-core - Well, than that is quite another story! :-)

http://www.pricewatch.com/cpu/

sidserious
Parent - By The Wizard (***) Date 2007-12-27 13:14
Hi,
    I am not seeing the problem either ... In the real world (not artificial benchmarks) these chips are better and far cheaper than the dual xeon setups .... 5Ghz when sub zero is easy enough, 4+Ghz on air depending how far you want to push the voltage ... They are easily the equal of 5365's a lot of people are useing in the engine rooms

The Wizard,Rybka 2.3.2a mp (2830) - Rechenschieber,Rybka 2.3.2g7 rmp 8 (2816) [B34]
Rated game, 3m + 0s Machines B, 16.12.2007

1.e4 c5 2.Nc3 Nc6 3.Nf3 g6 4.d4 cxd4 5.Nxd4 Bg7 6.Be3 Nf6 7.Nxc6 bxc6 8.e5 Ng8 9.f4 Nh6 10.Qf3 0-0 11.0-0-0 d5 12.h3 f6 13.Bc4 Bb7 14.exf6 Bxf6 15.Ne4 e5 16.Bc5 Be7 17.Qb3 Rb8 18.Bxa7 exf4 19.Bxb8 Qxb8 20.Be2 Qc7 21.h4 Rb8 22.h5 Ba6 23.Qc3 Bxe2 24.hxg6 f3 25.gxh7+ Kxh7 26.Qe3 Bf8 27.Ng5+ Kg6 28.Nxf3 Bxf3 29.gxf3 Qd6 30.Rdg1+ Kf7 31.Rh4 Rb4 32.Rxb4 Qxb4 33.Re1 Kg8 34.Qe6+ Kh8 35.Kb1 Bg7 36.c3 Qh4 37.Rd1 Qc4 38.Qg6 d4 39.cxd4 Qd5 40.Qe8+ Kh7 41.Qe4+ Nf5 42.a4 Qxe4+ 43.fxe4 Nxd4 44.a5 Nb5 45.a6 Be5 46.a7 Nxa7 47.Rd7+ Kg6 48.Rxa7 c5 49.Rd7 Bf6 50.Kc2 Bg7 51.Rxg7+ Kh5 52.Rg5+ Kxg5 53.Kd3 Kf4 54.Kc4 Kxe4 55.Kxc5 Ke5 56.b4 Ke6 57.b5 Ke7 58.b6 Kd7 59.Kd5 Kd8 60.Kd6 Kc8 61.Kc6 Kd8 62.b7 Ke7 63.b8Q Kf6 64.Kd5 Kf5 65.Qe5+ Kg4 66.Ke4 Kh3 67.Kf3 Kh4 68.Qd5 Kh3 69.Qh5# 1-0
Up Topic Rybka Support & Discussion / Rybka Discussion / Unintended consequences of Intels 45nm process

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill