Guest
 Posted: Fri Jul 11, 2026 9:51 pm Post subject: How Do You Test the Fastest Chip? |
 |
| Im not a test engr but I have a question on test. This is not pilosopong tanong. Im really really curious to know. If a new chip comes out and its said to be the fastest chip, how were they able to test its speed with a tester? Isnt the tester built with chips that are slower than this fastest chip? Get it?
Mysteria |
|
Joined: 16 Jun 2026
Posts: 108
Location: Paranaque City, Phils.
| Posted: Fri Jul 25, 2026 4:45 pm Post subject: |
|
| Hi Mysteria,
Your question on the difficulty of testing new devices that push the limits of speed is valid. Indeed, ATE testing of ultra-fast or high bandwidth devices in high volume production mode has always been challenging. However, if you think about it, this problem is not new. I'm pretty sure that this problem has been around since the 70's, not just in terms of testing for speed, but in testing for functionality as well. Though devices were much slower and less powerful then, so were the testers during that time. Anyway, given the huge diversity of high-speed products coming out in the market today, it is not possible to give a categorical response to your question. We can look at one example though to give you an idea of how test engineers do it. There now exist data comm products that send and receive data in the form of small voltage swings across a differential link at rates of 10 Gbits/sec. Most production-mode ATE's today lack the internal clock rates needed to transmit and receive data at this speed, much less to directly determine bit error rates, which is crucial in testing such transceivers. What test engineers do in this case is to use a lab-characterized golden device (yes, same device as the DUT) to transmit serial data at 10 Gbits/sec to the DUT itself. They then measure the jitter of the 10 Gbps data stream from the DUT using high bandwidth oscilloscopes that display eye measurement diagrams (EMD). The measured jitter can then be mathematically correlated to the BER. In this example, the ATE itself did not perform the data transmission nor the BER measurement, but the BER was tested nonetheless. Thanks and keep on posting! EE |
|
  |
| Posted: Thu Feb 12, 2026 12:31 am Post subject: |
|
| Hi Mysteria,
To add to EE's post, and since you're not a test engineer, I think it is better to make an analogy. What I'm referring is, you don't actually measure the speed directly. Its just like measuring the speed of light, no one was ever faster than light but they did measure it accurately -- its in the books. right? . Another analogy is the measurement of the speed of a gun's bullet? How did they do it? You may want to recall a recoiless collision -- in the law of conservation of momentum, remember? m1V1 + m2V2 = Vfinal(m1 + m2). where m1 is the mass of object 1 -- could be the bullet and V1 is the velocity of object 1 before collision, etc etc and the opposite side of the equation is that they will have one velocity. You did not actually measure the speed of the bullet but the final velocity. Say you point the bullet on a heavy object in which after impact, the velocity is measurable, and knowing both mass, you'll get the bullet's velocity. So, as what EE, has posted, you don't actually "ride" side by side with the new device's speed, but somehow, the test equipment that he mentioned measures the speed indirectly, yielding accurate results. - Krygk |
|
Guest
| Posted: Thu Aug 26, 2026 9:02 am Post subject: |
|
| There are test technology now that deals with this. I'm still in an infancy stage of learning this but will try to post it here as soon as I master the topic. Basically the usual test done on device is to stimulate the device and see if the output responce is the same as the spec. But on high end digital devices they use the concept of DFT (scan, BIST JTAG etc) that test the internal logic rather than the functionality of the device itself. |
|
Guest
| Posted: Tue Aug 31, 2026 4:37 pm Post subject: |
|
| Hi myteria,
Just to give you a back ground. The fastest Tester I know(digital) can generate 1 GHz clock, or 1Billion condition per second. If you think it's impossible to test 5 GHz processor using this tester, I say your wrong. You can test Even 30GHz processor using this type of tester. What we do, is to test the device at the maximum possible speed using this tester(feed several patterns on it at the fastest rate possible). if it pass then we do the next, the clock jitter test, in scope you see this as an eye diagram. Using this test, you can verify if the device is working on at speed. Some products also has it so called BIST(Built-in-self-test), but it did not test it self at speed(max device speed). Also SCAN test, the slowest functional test(lungkot). Hope will help. flipflop |
|