Why One Diamond Bit Needs Water and Another Cuts Granite Dry

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Before the technique, the premise needs fixing, because it is wrong in a way that will ruin a tool. A diamond bit does not need water. Some do and some do not, the difference is large, and it has nothing to do with the diamond. Our own catalogue carries the proof side by side: one diamond tool here is sold on grinding glass with water, and another is sold on cutting granite dry with no water at all. Same abrasive. Opposite instruction.

What separates them is the bond, meaning the stuff holding the grit on. And once you see that, the second half of the usual advice also turns out to be wrong: a diamond bit does not want a low speed either. It wants a particular speed at its cutting edge, and because edge speed is diameter times rotation rate, the smallest bit in our range and the largest differ by a hundred times in diameter, so a single instruction cannot possibly serve both.

A bit has no speed, and our own earlier page did this arithmetic first

This part is already settled on this site and I am not going to repeat it. Our page on why your drill bits go blunt so fast, and why the fix is to slow down and push harder works it out plainly: At a fixed 2000 rpm, the rim of a one eighth inch bit covers roughly 65 feet a minute. A half inch bit, same drill, same dial, same trigger, covers roughly 262 feet a minute. That is right, and it is the single most useful thing anyone can know about a drill dial.

What that page then says, and says deliberately, is that there are two rules and they are opposites. Slow and heavy in steel, because high speed steel loses its temper and the chip is what carries the heat away. Fast with hammer action and a light touch in masonry, because carbide crushes rather than shears and tolerates heat. Its closing argument is that using either rule on the other material is how both kinds of bit get wrecked.

Bonded abrasive in glass is a third regime, and it breaks both of those rules at once. It is slow like steel, light like masonry, and wet like neither. The reason is that nothing in it is cutting.

Grinding has no chip, so it has no cooling path

A twist drill in steel shears a chip off, and that chip walks out of the hole carrying most of the heat with it. That is the mechanism our earlier page is built on, and it is why cutting fluid there is a helpful extra rather than the whole story.

A diamond bit removes material by abrasion. Thousands of tiny hard points scratch microscopic amounts away, so the debris is powder, and powder has an enormous amount of surface and almost no mass. Put a number on it. Grind a 1 mm hole 3 mm deep in glass and you have removed 2.36 cubic millimetres, which at the density of glass is 5.9 milligrams of swarf. Suppose every last grain of it leaves the hole 300 kelvin hotter than it arrived, which is generous. All of it together carries away about 1.5 joules. A rotary tool putting a modest 10 watts into that contact for one minute has delivered 600 joules. So the debris removes a quarter of one percent of the energy and the other 99.75 percent stays in the workpiece and the tool.

That is the real content of the water advice. In steel, fluid helps a cooling path that already exists. In grinding there is no cooling path, so the water is not assisting the mechanism, it is the mechanism. Which also tells you how much you need, and the answer is less than people expect. Water absorbs 83.7 joules per gram for a 20 kelvin rise, so carrying 10 watts away needs 0.12 grams a second, which is about 7 millilitres a minute, or a teaspoon and a half. That is a drip, or a shallow bath, and it is exactly the technique everyone recommends without saying why it works.

The same arithmetic also says why the big tool cannot be managed the same way. A 200 watt cut needs 143 millilitres a minute, twenty times as much, which is a flow and not a drip. And it explains the one instruction everybody gives and nobody justifies, which is that the water must stay liquid. Boiling absorbs 2,260 joules per gram against 83.7 for warming, a factor of 27, but a surface that has gone dry absorbs nothing at all. Steam is not cooling, it is the sound of the cooling having stopped.

The bond is the whole story, and it inverts the failure diagnosis

Diamond is the hardest material in the hole by a wide margin. It is not going to be blunted by glass. So when a diamond bit stops working, something else has happened, and all three of our records describe it in the same slightly misleading way.

A 24 piece diamond drill bit set for glass, 1 mm to 2.5 mm with a Dremel fit is the clearest case and the one whose figures the drawing above is built on. Its record publishes the four sizes, 1 mm, 1.5 mm, 2 mm and 2.5 mm, on 1/8 inch shanks with four brass collets, and it says the bits grind rather than cut. It then says, twice, that the coating is the wearing part of each bit, and lists as a drawback that diamond tips wear over time.

That is the framing worth correcting. An electroplated bit holds its grit in a thin plated layer that the stones stand proud of, so each one is gripped only around its base. Heat arriving at the grit has nowhere to go but through that thin layer. Let the layer get hot enough and it loses the grip, and the diamonds leave while still perfectly sharp. The tool has not gone blunt. It has gone bald. The practical difference matters, because a blunt tool rewards more pressure and a bald one does not: pressing harder on a bit that has lost its grit just rubs steel on glass, which is the fastest way to crack the piece and the slowest way to make a hole. Our 24 piece glass drill bit set review has the rest of the record.

A Drilax Pro 4 inch diamond hole saw is the other end of the axis and it is the product that disproves the premise. Its record leads on vacuum brazed industrial diamonds, claims rapid dry drilling with no water needed, carries a comparison row that reads simply Water: Dry drilling, and is rated at over 100 holes. In a brazed bond the alloy wets each stone and grips it around most of its surface rather than at its base, so the grit is held mechanically and chemically, and the thick brazed layer is itself a wide path for heat to leave by. That is what buys the dry capability. It is not that the diamond is better. It is that losing the grip is much harder.

And then the hedge, which is the honest part of that record and worth quoting because it is doing real work: Although it drills dry, very hard or prolonged cuts run cooler and cleaner with some water. Read that next to the mechanism and it is precisely right. Dry is available because the bond tolerates heat, not because heat has stopped being generated, and the longer the cut the more of it accumulates. Our Drilax Pro 4 inch diamond hole saw review keeps the safety notes that matter more than any of this.

A Drilax 30 piece diamond drill bit set in 2 mm and 3 mm sits on the coated side with the first one and is the most explicit of the three about it. Its record describes a triple layer reinforced diamond coating, says plainly that the coating does the grinding so the bits abrade rather than cut, and its own not-for field reads: fast drilling or dry, uncooled use. That is the same tool class as the 24 piece set and the same instruction, and the two records together against the hole saw's record are the entire argument of this page, sitting in our own database. Our Drilax 30 piece diamond bit set review covers the sizes.

Feed rate, which is the half nobody writes down

Speed gets all the attention and feed gets none, and feed is where beginners actually fail. There are two ways to do it wrong and they look nothing alike.

Too little pressure and the grit skates instead of biting. A diamond point that is not penetrating is still rubbing, so it is still making heat, and now it is making heat without removing material. Our records call the result slow progress and a worn coating; the mechanism is that the tool is polishing the glass and the glass is polishing the tool. This is the failure that the advice to press lightly accidentally causes, and it is why both small-bit records pair light pressure with water rather than offering it alone.

Too much pressure and you are no longer limited by the tool at all. Glass fails in tension at a crack, and a point load on a brittle plate puts tension underneath the contact and around the far face, which is why every one of our guides says to ease off on breakthrough. The thinnest remaining material is also the least supported, so the last fraction of a millimetre is where a piece that drilled perfectly for two minutes splits. Our page on how to drill glass and gemstones without cracking them gives the right technique for all of this, and gives it well. It is correct on every point. What it does not do is say which failure each instruction is defending against, and the two failures want opposite corrections.

The window between them is narrow and it has no published width. Not one of our listings publishes a feed force, a feed rate, or a recommended speed. Our page on how to choose a diamond core drill bit for tile and stone is right that you should decide dry or wet and respect the material limits, and it is right that percussion strips the diamond bond, which is the same mechanism this page is about. It too has no numbers available to it, because there are none.

The honest limit, and the free technique

The limit is severe and it is worth stating bluntly. Across every diamond product we carry, the number of published figures that would let you set a dial correctly is zero. No surface speed, no RPM range, no RPM ceiling, no feed force. The drawing above uses the spindle speeds a reader would plausibly select, not recommended ones, because none are recommended.

One consequence of that gap is sharp enough to call a defect. The 4 inch hole saw's record supports two mounts: an angle grinder through a 5/8 inch by 11 thread, and a drill through an adapter. Those are not two ways of doing the same thing. A grinder commonly turns at around 11,000 revolutions a minute and a drill at a few hundred, so the same saw on the two mounts its own record recommends runs its edge at 58.5 metres per second or at 2.66, a factor of 22, and the record publishes no limit for either. A buyer choosing between the two mounts on convenience has no way to know they are choosing between two different processes.

The other consequence is the one that finally makes the range make sense. A 2.5 mm bit at 20,000 revolutions a minute runs its edge at 2.62 metres per second. The 4 inch saw at 500 revolutions a minute runs its edge at 2.66. Those are the same number. A hundredfold span of diameter and a fortyfold span of spindle speed cancel almost exactly, which is the clearest possible demonstration that the edge is the thing with the speed. Going the other way, the 1 mm bit would need 50,800 revolutions a minute to match, and rotary tools stop near 35,000, so the smallest bit we sell cannot reach the edge speed of the largest even at full throttle. That is not a flaw. It is the reason small bits feel so slow, and it is unfixable by buying anything.

So the free technique, and for once it is free in the strongest sense: the thing that most extends the life of a bonded abrasive costs nothing and is not in any of our guides. Keep the contact moving. A diamond point held in one spot loads its gaps with the powder it just made, and once the gaps are full the grit cannot reach fresh material and the tool switches from abrading to rubbing, which is the runaway our metal page describes arriving by a different road. Lift the bit clear every few seconds, let the water flush the face, and go back in. On the small bits this is the difference between a bit that does one hole and a bit that does twenty, and it is the real reason a shallow water bath outperforms a careful drip: the bath flushes, and a drip only cools.

Two other free habits follow from the same mechanism. Start the hole at a tilt and bring it upright, which both our guides already say, because a tilted start gives the grit an edge to bite rather than a flat face to polish. And drill from both sides on anything you can reach the back of, stopping before breakthrough and finishing from the far face, which removes the unsupported-last-fraction failure entirely rather than managing it.