[Help]Intel I7 920+Sapphire Pure Black X58

[whoololon]

I hope this helps.
The definitive "How to overlock the i7" thread - Chilled PC Forums
Normally, the values that are not displayed can be assumed to be set to "AUTO", but you are right to ask.

[fjavi]

@Jeff_lindsay:

Hello again, the thing is that I think you're not going about it the right way. I haven't touched the Multiplier, which according to what I've read is the "CPU Ratio" on my motherboard, you can see in the first screenshot that it's at 20, and it won't let me touch it, I suppose this micro has the multiplier locked.
I also haven't touched the CPU VCore, it's on Auto, the only thing I touched was the Memory Voltage, which if I'm not mistaken is the "DIMM Voltage" on my motherboard, and I set it to 1.6v because the sticker on my memory says it's 1.6v.

What I still don't understand is that 0.5 difference that I've read about in several guides, and I don't want to fry anything. I refer to my previous messages so you can see my doubt, even by putting everything on auto those 0.5v of difference don't appear, and although I want to set the CPU VTT manually the minimum is 1.2v, when I've read in guides that it should be 1.1v. I don't know if you understand what I mean.

Let's see if someone can resolve this doubt.

Regards!

That micro doesn't have the multiplier locked, it depends on the motherboard, Asus and some others allow using 21, but it's up to each manufacturer of the boards, I have two Asus and they allow 21 on the 920, but I had a DFI X58 that only allowed 20, using the same CPU in both brands.

I don't pay any attention to that 0.5 difference, the CPU VTT is voltage for the CPU's memory controller, you'll have to put a voltage according to the frequency and latency of the memories, for 1600 MHz and 9-9-9 I usually put 1.267v, although this will depend on each board and each CPU because some controllers are better than others, that's with the same CPU.

If you use for example a 200x20 and that CPU is D0 you should put a Vcore of 1.25v although this is according to the board and according to the CPU, some hold with less voltage and others ask for more, but only vcore.

You should only focus on Vcore example:
Vcore 1,256v
memory frequency 1600
CPU VTT up to 1,287v that's at 9-9-9 if you lower latency it's easy to need more
Uncore at double the memory, that is 3208 I think you'll get.
IOH vcore 1,20v
Memory voltage to 1,60v
CPU PLL vcore: 1,88 to 1,90v
If you disable HT it's easy to ask for less voltage and get less hot therefore be more stable.
I think those are the key options for that socket 1366, to tweak would be to raise something CPU vcore and CPU VTT the rest doesn't need to be touched, leave loadline control at 100%.

You shouldn't touch much more.

The CPU Multiplier you will be able to lower, I suppose using the numeric keypad the + and - minus would be to lower and plus to raise.

Regards

[Jeff_lindsay]

@whoololon:

Espero que ésto te ayude.
The definitive "How to overlock the i7" thread - Chilled PC Forums
Normalmente, los valores que no se reflejan puede sobrentenderse que van en "AUTO", pero haces bien en preguntar.

Gracias por la guía whoololon, ya he aclarado todos los conceptos perfectamente.

@fjavi:

Ese micro no es que tenga el multiplicador capado es que depende de la placa, las asus y algunas otras permiten usar el 21, pero es cosa de cada fabricante de placas, yo tengo dos Asus y permiten el 21 en el 920, pero tenia una DFI X58 que solo permitia el 20, eso usando en ambas marcas el mismo CPU.

Yo no hago ni caso a esa diferencia de 0,5, el CPU VTT es voltaje para el controlador de memoria del CPU, tendras que meter un voltaje según la frecuencia y latencia de memorias, para 1600 MHz y 9-9-9 yo suelo meter 1,267v, aunque esto ya dependerá de cada placa y cada CPU pues algunos controladores son mejor que otros, eso en CPU iguales.

Si usas por ejemplo un 200x20 y ese CPU es D0 deberías meter un Vcore de 1,25v aunque esto es según la placa y según el CPU, unos aguantan con menos voltaje y otros piden mas, pero solo vcore.

Te debes fijar solo en Vcore ejemplo:
Vcore 1,256v
frecuencia de memoria 1600
CPU VTT hasta 1,287v eso a 9-9-9 si se baja latencia fácil que necesite mas
Uncore al doble que la memoria osea 3208 creo que te saldrá.
IOH vcore 1,20v
Voltaje de memoria a 1,60v
CPU PLL vcore: 1,88 a 1,90v
Si desactivas el HT fácil que pida menos voltaje y se caliente menos por lo tanto sea mas estable.
Yo creo que esas son las opciones claves para ese socket 1366, de retocar seria subir algo CPU vcore y CPU VTT lo demás no hace falta tocarlo, dejar loadline control a 100%.

poco mas debes tocar.

El Multiplicador de CPU si lo podras bajar, supongo que usando en el teclado numérico el + y el - menos seria para bajar y mas para subir.

saludos

Gracias a ti también fjavi por ayudarme con este OC, de verdad que os estoy muy agradecido.

Ahora paso a revelaros mi primera experiencia que cuanto menos es un poco extraña.

Después de fijar bien los conceptos de Voltajes, y estrategia a seguir para hacer el OC, comencé con estos voltajes, y el PC no arrancaba, ni siquiera aparecía la info la BIOS, sin mas se quedaba la pantalla en negro.

Aquí os dejo como comencé:
CPU Vcore 1.25v
DIMM Voltage 1.6v
CPU VTT 1.25v
IOH Vcore 1.20v
QPI PLL 1.27500v
IOH/ICH IO Voltage 1.500v
CPU PLL Vcore 1.88v
ICH Vcore 1.10

Con estos valores no arrancaba el PC, probé a subir algunos valores un poco, y quedó así:

CPU Vcore 1.32500v
DIMM Voltage 1.65v
CPU VTT 1.37500v
IOH Vcore 1.20v
QPI PLL 1.35000v
IOH/ICH IO Voltage 1.500v
CPU PLL Vcore 1.88v
ICH Vcore 1.10

Nada tampoco arrancaba, pantalla en negro y no cargaba ni la BIOS, la solución era CMOS y reiniciar sino ni reiniciaba la placa.

Después de esto me propuse a dejar algunas cosillas en Auto, para intentar ver cual era el valor por el que no arrancaba.

Y bueno, con estos valores arranca perfectamente:

CPU Vcore 1.20000v
DIMM Voltage 1.60v
CPU VTT 1.20000v
IOH Vcore [Auto]1.100v
QPI PLL [Auto]1.100v
IOH/ICH IO Voltage [Auto]1.500v
CPU PLL Vcore [Auto]1.800v
ICH Vcore [Auto]1.050v

Ahora mismo estoy con esos valores testeando:

Target CPU Frequency: 4000 Mhz (200x20)
Target Memory Frequency: 1600 Mhz (2:8)
CPU Frequency: 200
PCI Frequency: 100
CPU Ratio: 20
Memory Frequency: [DDR-1066 (2:8)]
Memory Timing: [By DDR-1600] 9-9-9-24
CPU Uncore Frequency: [2400 MHz (18x)]

Le estoy pasando el Prime 95 en modo Blend, creeis que debería hacerle algún otro test?

Saludos y gracias.

[Jeff_lindsay]

Hello again, I have already managed to make the OC stable, below I put the values so that you can comment on what you think:

Target CPU Frequency: 4000 Mhz (200x20)
Target memory Frequency: 1600 Mhz (2:8)
CPU Frequency: 200
PCI Frequency: 100
CPU Ratio: 20
Memory Frequency: [DDR-1066 (2:8)]
Memory Timing: [By DDR-1600] 9-9-9-24
CPU Uncore Frequency [2400 MHZ (18x)]

CPU Vcore: 1.20625v
DIMM Voltage: 1.60v
CPU VTT: 1.25v
IOH VCore: 1.100v
QPI PLL: 1.150v
IOH/ICH IO Voltage: 1.500v
CPU PLL VCore: 1.800v
ICH VCore: 1.050v

I did the stability tests with Prime 95 in Blend mode for 9 hours, LinX with 20 passes with memory in ALL, and zero errors with the previous values.

Until I reached those values I had several errors including system restart, system freeze, BSOD 0x124 and BSOD 0x1E. By touching VCore, CPU VTT and QPI PLL I achieved stability. Depending on the error, I had to adjust one thing or another, for this I took a good guide.

The temperatures after 9 hours of Prime 95 were 64º as a maximum. What do you think?

Greetings and thanks for the help!

[Jeff_lindsay]

Someone could comment on what they think about the voltages and temperatures?…It's my first OC, and I wanted to know if everything is okay.
Please...:cry:

[Sylver]

Everything is very good, very correct values. Congratulations, your overclock is quite satisfactory

Best regards!

[Jeff_lindsay]

@Sylver:

It's all very good, very correct values. Congratulations, your overclock is quite satisfactory

Best regards!

Thanks Sylver, to be honest this is my first contact with the world of O.C, and that's why I ask for the opinion of someone who is an expert.

I don't want to seem pushy, but... Do you think it could have been much better?, What could I have improved?, Voltages?, Frequencies?...

Best regards and thanks for the help Sylver!

[Sylver]

Nothing man, it's very simple, you've already seen how it goes. You know the theory and apply it correctly. You keep increasing the frequency, you keep passing stress tests until it's not stable, that's the limit. If you increase the voltage minimally you can check if it goes up a little more, but it's very likely that temperatures will increase, and you may not achieve stability, that's normal.

In total, as long as it doesn't reach 70º degrees or pass it, it's fine. At that temperature under full load it has reached as far as it can go. It's so random that even 3 micros identical to yours can reach values with up to 1/4 difference between them, both in frequencies and temperatures. So you know, it's not an exact science in any of its aspects

¡Greetings!

[goodlucky1234]

¡Esta publicación está eliminada!

[Jeff_lindsay]

@Sylver:

Nothing man, it's very simple, you've already seen how it goes. You know the theory and apply it correctly. You keep increasing the frequency, you keep passing stress tests until it's not stable, that's the limit. If you increase the voltage minimally you can check if it goes up a little more, but it's very likely that temperatures will increase, and you may not achieve stability, that's normal.

In total, as long as it doesn't reach 70º degrees or pass it, it's fine. At that temperature under full load it has reached as far as it can go. It's so random that even 3 micros identical to yours can reach values with up to 1/4 of a difference between them, both in frequencies and temperatures. So you know, it's not an exact science in any of its aspects

Best regards!

Well many thanks Sylver, maybe next year I'll dare to raise it to 4.2 Ghz, but I don't know if it will be worth much the truth. From 4.0 to 4.2 Ghz I don't think the gain will be much, I speak from ignorance of course.

Best regards!