夏天到了,大家的室温很可能都已经接近或者突破了30度大关
而同样有很多同学用着四核或者开了四核的AMD CPU并且超频使用
而散热器又不见得很好
所以到了夏天,玩游戏或者进行负载较大的工作时,可能会发生不稳定现象
别担心,绝大多数情况下,这不是你的U缩缸了
而是AMD的热墙在影响你CPU的稳定性
下面就来说说什么是热墙。
前言
我们知道晶体管在高温的时候会发生电子迁移,造成信号的噪音加大(我不是学这个的,所以对具体原因不了解,我们也不必深究,知道结果即可),从而对超频造成不稳定因素增加,表现就是,你在达到某个频率后,要想再提升一点点,就要大幅加压,而在大幅加压的同时,发热也大幅增加了。
热墙从何而来
AMD的热墙最早可追溯至K8早期,只要在温度“不太高”的时候,超频的稳定性就增加许多,而当温度达到55度以上时,你就会发现在某个电压和频率下,原来可以通过稳定性测试、或者跑分测试的,现在却不能了。到K10的时候,这种现象尤为明显,在温度达到55度以上时,CPU负载加大的时候,CPU主频或者NB频率过高而引起的不稳定现象大幅增加。最直接的表现就是你看到蓝屏代码124或者50。这种现象在超频到3.8GHz以上、主板只有4+1或者更弱的供电时,尤为明显。
在这里我们把Intel的CPU拿来做比较,撇开测温准不准的问题,i7在4GHz下烧机温度基本都在70度以上,而体质或散热条件不太好的时候甚至可以高达90度,但是照样可以稳定工作,AMD的就不行。这可以把AMD的CPU比作一个比较大的电阻(Low-K),而Intel的电阻比较小(High-K),这样就使得AMD的CPU电流较小,对信号的稳定性控制较差,这就形成了热墙,但是在频率提高的时候发热量不会大幅暴增。而Intel则相反,电流较大,对信号稳定性有较大帮助,而带来的问题就是频率提高后漏电现象大幅增加,温度也大幅提高。
如何“推”热墙
热墙是个讨厌的东西,在冬天可能没什么,室温低的时候CPU不会那么容易达到55度以上,而在夏天的时候,你使用的是四核,又使用电压1.4V或以上超频到3.8GHz(NB2.6GHz)或更高时,哪怕你用最高端的风冷散热器,都很难在室温30度的时候把温度控制在55度以下(别跟我说你才40来度,那是主板测温不准,技嘉7系主板全系列通病)。所以,在这个时候,热墙永远会影响你,你不可能把热墙推得一干二净,但是你可以尽量减少热墙的影响。首先,你要搞清楚到底是主频还是NB频率不稳定,一般可以用Prime95来检测,如果是NB频率不稳定,大概有以下几种情况:
1. 秒杀,或者直接蓝屏,建议你直接降至少50MHz的NB频率再说。
2. 在7分钟以内红一个核心,说明你还要再加0.025V的NB电压。
3. 如果20多分钟红一个核心,那么再加0.0125一般就可以稳定。
4. 对于某些开核不稳定的U,体质比较弱的核心在这个时候可能会红。
5. 有说法是“DDR2平台比DDR3更容易撞热墙”。
如果是CPU主频不稳定,一般你会直接看到蓝屏代码124或者直接花屏重启。解决办法有以下几种:
1. 换更好的CPU散热,或者改进你的机箱风道。
2. 想办法降低室温,开空调。
3. 在散热OK的前提下,多加两档核心电压。
4. 提高NBPLL电压或许会有帮助(1.8V的那个)。
5. 如果你跑P95坚持了30分钟以上挂了,不用管他,平时用你挂不了。
6. 降频。
AMD对“热墙”的改进
Thuban核心现在已经显得更能在热墙中稳定,大概在64度的时候还可以坚持在4G-4.2G下。而GF更是透露在32nm的时候会采用high-K工艺,这样应该可以提高信号的稳定程度,减少热墙的影响,这一切都要等到推土机的超频测试公布后才会有结果。
以下是全世界最早最详细的K10超频心得,来自Dolk大神,这里节选其中他对热墙的体会,我就不翻译了:
The Heat Wall
Throughout the better part of my overclocking I have repeatedly hit the same wall over and over again, and never realized what was going. This wall is what I like to call the Heat Wall. It is pretty simple: the Phenom IIs love to be cold, just like the Opterons from the socket 939. You go any higher than 55C on your CPU, usually you will have problems with keeping your OS stable. Now that the Phenom IIs have no cold bug, these new CPUs seem to has taken a liking to the cold more so than before.
To explain all of this in detail, I am going to bring you on a long journey through my frustration at trying to achieve 4.0 GHz on a 32 bit OS. From the very start of owning the 940BE I have constantly been trying to get 4.0 GHz, which should have been achievable with my setup. I have a decent water-cooling, RAM along with a great motherboard. But no matter what I did, I could only boot up for about 5 minutes, or instantly crash, or repeated BSODs. I have seen many people get to the 4.0 GHz using less efficient cooling, some using air setups. My guess is that they received a really good batch of CPUs, which happens. But for the rest of us, there was no way to get to the 4.0 GHz range without using extreme cooling. I finally lugged my computer outside to the freezing cold where I saw cold boot ups close to 5C. And what do you know, I could boot up into 4.0ghz and do whatever I wanted to do. After trying to go higher, I found that I could only go as high as 4.2 GHz. I tried this same process a couple days later but this time the temperature outside was a bit warmer so I did not get the same crazy low temperatures. This time I could only get up to 4.1 GHz. I tried the same settings I did before and there was no luck. I concluded my work saying that the computer must be colder.
When the 720BE came out I started to get data on what it could do as far as overclock. The data seemed to be the same and followed most of my rules, but there were those that went outside of everything. The 720BE data that was being ran on the AM3 socket, resulted in higher overclock with less voltage. But they still had a limit even if it was 35% above stock. This limit was the same I experienced with my 940BE. In which, no matter what settings you used and even though the CPU was only using 1.45v, you could not go any further without having to resort to extreme cooling.
Thus, there is a limit to the Phenom II at ‘X’ CPU temperature. No matter what you do, you will not be able to go any further with your CPU when it hits this wall. You should be able to notice when it hits this wall on a AM2+ system. Because normally your CPU should be somewhere near a load temperature of 55C. On an AM3 system, it is a bit more difficult to notice, but there is a general rule that you can use. If your CPU is not on extreme cooling, if your CPU has hit close to25-35% increase, and your load temperatures are below 50C but above 30C than you have most likely hit the wall.
My only suggestion to aground this wall, is to get better cooling. The main idea behind this section was to give you an understanding that there is a wall, and no matter what you do, you will not be able to go around it. I have seen many cases where people complain that even with no heave voltage and low temps, no matter what they do they cannot go further, unless they resort to extreme cooling. There may be another way but so far I have yet to see one.
转载PCEVA作者royalk