It's fixed to the cupola but the cupola can rotate independent of the turret.
Way too much information follows:
The commander's main means of surveying the battlefield is a forward-facing TKN-3M pseudo-binocular periscope, augmented by two rectangular TNPO-160 periscopes on either side and two TNPA-65A periscopes embedded in his hatch. The TKN-3 periscope is aimed directly forward and is aligned with the centerline axis of the cupola. The two TNPO-160 periscopes are oriented 45 degrees from the centerline of the cupola.
With just these two periscopes, the commander has vision in a 176-degree frontal arc with a blind spot of 22 degrees to the direct front which is filled by the TKN-3. Because the cupola can rotate, the five periscopes in the cupola provide the commander with an all-round view when he is buttoned up. There is no periscope that allows the commander to see directly behind the turret. For that, he must spin the cupola to one side and look out of either one of his TNPA-65A periscopes, although the anti-aircraft machine gun would usually be in the way as it is stowed directly behind the cupola in the 'travel' position when not in use. Due to the conformal slant of the gunner's hatch on the left side of the turret, the commander's view to the left of the turret is largely unimpeded. Even the gunner's night vision sight does not completely block the commander's view as the height of the sight housing does not exceed the maximum height of the turret roof.
However, the commander's view to the left of the turret in the 10 o'clock sector was obstructed when Kontakt-1 reactive armour blocks were installed on the roof of the turret beginning with the T-72AV modification in 1985. This problem persisted when Kontakt-5 blocks replaced the Kontakt-1 blocks in the T-72B obr. 1989 model and continues to plague the T-90A. For these later models, the burden of monitoring these sectors falls upon the gunner.
For general vision, the commander is provided with four periscopes to supplement the TKN-3. In total, the field of view of the commander from the cupola (without head movement) is 288 degrees, with a 72-degree dead zone to the rear.
The commander's cupola of the T-64 lacked an anti-aircraft machine gun and was furnished with only one TKN-3M periscope and two TNPO-160 periscopes. The field of view (without head movement) was 144 degrees. This cupola was carried over to the T-64A. Given that a successful template for a periscope layout in a cupola of this design was already established since the T-54 obr. 1949, it is a mystery why the T-64 cupola had such constricted visibility. Needless to say, the T-72 was vastly superior in this particular aspect. In 1975, a new and much more technically advanced cupola with a ZU-64A remotely controlled anti-aircraft machine gun system was implemented on the T-64A obr. 1975. Two TNPA-65 periscopes were finally added to the hatch of the new cupola, but to accommodate the PZU-5 sight for the ZU-64A system, the TNPO-160 periscope on the left of the TKN-3 had to be removed. As a result, the commander's visibility was still not on par with his T-72 contemporary. In fact, the higher statistical weight of forward-facing periscopes compared to side or rear-view periscopes makes the new cupola a downgrade over the older version, despite the increase in the number of periscopes. These nuances are important when evaluating the validity of various cupola designs.
Compared to a typical Western tank cupola, the number of fixed periscopes in the T-72 model is clearly less, but the number alone is not necessarily indicative of actual utility. For example, the Leopard 1 provided its commander with eight periscopes arranged around his circular cupola, but only five are aimed in the forward 180-degree sector and two of them are partly obstructed by the loader's cupola, loader's machine gun skate mount and loader's hatch opening mechanism on the left side of the turret. It is also important to note that the commander's cupola on the Leopard 1 does not rotate and the forward-facing periscope has a very high periscopicity so that the field of view is inherently narrower. In other words, the number of vision devices providing a view towards the forward half of the turret is not more than in the T-72 commander's cupola and there are other secondary factors that affect the commander's visibility. A T-72 commander only loses out in convenience when directing the driver to reverse the tank as he must rotate his cupola in order to see behind the turret or have the turret aimed to the rear.
To further expand our perspective, it should be noted that the commander of an M60A1 is furnished with eight M41 prismatic vision blocks arranged around his oblong M19 cupola, with one aimed forward to cover the 11 o'clock sector, two of them aimed in the forward arc to cover the 10 o'clock and 2 o'clock sectors, two of them aimed to the sides, and three of them aimed in a 7 o'clock to 5 o'clock arc. There is one wide-vision periscope installed just behind and above the M85 machine gun in the cupola and aimed directly forward. Adding on the fact that the M19 cupola can rotate, it is clear that an M60A1 commander has much better visibility than a T-72 commander under practically all circumstances. However, none of the M41 vision blocks are heated, so fogging will tend to seriously degrade visibility in chilly weather. Also, the objectively poorer rearward visibility from the T-72 cupola compared to Western tanks does not necessarily translate into objectively poorer combat performance as the value of observation devices depends on the context in which they would be used. It is a perfectly valid observation that when the tank needs to reverse, it is often in a non-combat situation where it is safe for the commander to be outside his hatch. In combat, it may be necessary to reverse in order to change positions or to reverse into turret defilade after firing a shot. In both cases - and in general - the driver would have approached the firing position from behind in the first place so he already knows that the area behind the tank is clear of obstructions and that he can freely reverse without fear of running into obstacles. If it is truly necessary for the commander to direct the driver when reversing the tank, the commander can rotate the cupola and use one of his periscopes for the task or open his hatch and peer out.
Furthermore, the fundamental purpose of the fixed periscopes has to be understood in order to assign them with their proper value. In combat, such periscopes are generally only useful if the enemy is very close to the tank (500 meters or less). Otherwise, they are only good for viewing the surrounding environment in order for the commander to gain a sense of spatial control over the tank, and this is done by finding landmarks. When the tank is moving speedily across rough terrain, observation through fixed unmagnified periscopes becomes ineffective due to the oscillation of the tank and the restricted field of view. The commander only sees an oscillating flicker between the ground and the sky, with no possibility of reliably discerning camouflaged enemy forces let alone identifying them.
For a modern tank created and fielded during the mid to late Cold War era, it is only practical to see and identify targets using a magnified optic and some form of stabilization is mandatory to allow it to be used effectively in a moving tank, as the narrower field of view through a magnified optic will exacerbate the negative effects of the oscillation of the tank. The TKN-3 periscope for the commander of a T-72 fulfills this purpose as it has a reasonably high magnification with a reasonably large field of view, and it has handles to allow the commander to hold it steady.
The characteristics of a tank commander's observation practices when buttoned-up in a fixed cupola with eight periscopes and one fixed forward-facing sight in the turret were examined in the 1974 study "Некоторые Статистические Характеристики Процесса Наблюдения Командира Танка" (Some Statistical Characteristics of a Tank Commander's Observation Processes) by G.G Golub et al. Three special cupolas were constructed to replace the original commander's cupola of a T-64 that was used as the experimental platform. The frequency and duration of usage of each of the viewing devices was recorded using a small forward-facing lamp on the commander's headset which would illuminate an array of photodiodes (light sensors) placed on top of each viewing device when the commander looks through the viewfinder. The first cupola design was a fixed type with eight fixed and equally spaced unmagnified periscopes arranged radially around the circumference of the cupola and one forward-facing TPD optic (modified periscopic sight with optical rangefinder removed). The second cupola design was the same as the first design but it had a stabilized electric drive for cupola rotation. The third cupola design was a manually-rotating type analogous to the T-72 cupola, having a total viewing arc of 206 degrees (± 103 degrees from the centerline axis of the cupola).
These cupolas were tested in various simulated combat conditions. The simulations were carried out in field conditions with moderately hilly terrain partly covered with bushes and trees. The targets included four tanks showing their frontal projection, three tanks in hull-down positions, two armoured personnel carriers, three ATGM teams, five recoilless rifles, and five anti-tank guns. All of these were arranged in such a way as to ensure that they would be uniformly concealed from the tank commanders as the tanks traveled down the pre-planned routes from a full 360-degree arc and at distances of 0.5 to 1.5 kilometers. The positions of the targets were shuffled throughout the experiments.
It was found that in general, 30% of all battlefield observations were carried out using the forward-facing unmagnified periscope and at most, 5% of observations were done using the magnified 8x optic with a stabilized field of view. However, it was also found that in tactical situations such as carrying out a breakthrough mission, the frequency of the use of a magnified optic to search for targets increases up to 50%. Overall, more than 70% of observations were made using only three periscopes at the front of the cupola covering a 100-degree frontal sector and over 95% of observations were made in a 200-degree frontal sector. Most interestingly, the experiments revealed that the highest recorded frequency of usage of the rear-view periscope was only 0.8%. It was also noted that the periscopes installed at more than 110 degrees off the centerline axis of the cupola (8 o'clock) were difficult to use due to neck strain when the tank was in motion. This was most likely why the commander's cupola of the T-80 used a rear-view prism embedded in the roof of the commander's hatch instead of a conventional periscope placed behind the commander's head.
Based on these results, it can be seen that in a fixed cupola with all-round visibility, five unmagnified periscopes covering the front 180-degree sector provide 95.3% of the total visibility needs of the commander under various combat conditions. The rear-facing periscopes are rarely used. A rotating cupola that provides vision in a 206-degree arc will fulfill 98.1% of the commander's visibility needs under the same combat conditions. In other words, the practicality of the T-72 cupola design can be considered to be experimentally validated. Even a T-64 cupola with just one TKN-3 and two TNPO-160 periscopes can theoretically fulfill 70% of the visibility needs of its commander, but on the other hand, the improved visibility from the two additional TNPA-65A periscopes in the T-64A obr. 1975 or T-64B cupola is offset by the loss of one TNPO-160 periscope.
Of course, the configuration of observation devices in the T-72 commander's cupola is certainly not perfect. A panoramic sight is ergonomically superior as the user's head does not need to move when the sight head rotates. The Leopard 1 is exemplary in this regard as it provided its commander with the excellent TRP-2A panoramic sight featuring a variable magnification of 4x to 20x, and beginning with the Leopard 1A4 in 1974, the commander was provided with the advanced PERI-R12 stabilized panoramic sight with a variable magnification of 2x or 8x. Panoramic sights were developed in the USSR during the 1930's and the PT-1 sight was the first to enter service, being installed on the T-26. Later, the PT-K panoramic sights were used on various modifications of the KV-1 and T-34, and indeed, Soviet engineers in the prewar era saw much greater value in panoramic observation devices compared to cupolas with multiple vision slits or periscopes and favored devices like the MK-4 rotating periscope (Gundlach periscope) and PT-1 for all-round visibility, but for one reason or another, postwar Soviet tanks were no longer equipped with such devices. Instead, all postwar Soviet armoured vehicles built in the 1950's standardized on the binocular TPK and TPKU periscopes paired with the TKN-1 night vision periscope, and beginning in the early 1960's, the TKN-3 series of combined periscopes became the new standard.