Archive for the ‘Tests and Reviews’ Category

Pushing the Gridseed 5-chip ASICs to the Maximum with a Voltmod

7, Apr 2014

gridseed-68-kohm-r52-resistor-mod

The last time we talked about voltmodding the 5-chip Gridseed ASIC devices we did what we considered the optimal voltmod for optimal performance/power consumption. Back then we discussed what could be the theoretical maximum you might be able to push the voltmod to in order to try to get the best possible overclock, though that would be achieved with an unreasonably high power usage. Well, we have decided to try out and push things to the maximum, namely replacing the R52 resistor with a 68 kOhm 402 1% SMD resistor. With this modification we were aiming to get 1400 MHz while still fitting in the power consumption that the ASIC could handle in Dual Mining mode or about 60W, however running in Scrypt only mode. With these settings we were planning to get stable around 600 KHS if we would succeed in actually running the GC3355 chips at a bit over 2 volts.

gridseed-68-kohm-mod-power-usage

So we’ve replaced the R52 with a 68 kOhm resistor, assembled everything up (do not try to run the ASIC disassembled with this mod as you can burn the voltage regulators in a matter of a few seconds). Firing up the ASIC has shown a total power usage of about 58-61W, so pretty much what we have planned initially, however the problem was that the Gridseed device did not want to function properly at that voltage after the modification. It simply did not send any shares back to the server, neither it was showing any HW errors, through apparently the device was working and communicating with the miner software. so we do not recommend to push things so much trying to get the maximum performance regardless of the power usage, better stay at at the reasonable recommendation by using a 47 kOHm resistor which should get you close to 500 KHS at about 20W, running the miners stable at about 1150-1163 MHz. Even if we were able to get the device working properly with the 68 kOhm resistor mod and with 60W power usage it wouldn’t be reasonable to get just about 100 KHS more with an extra 40W of power. You could of course try resistors with values in between 47 and 68 kOhm to see where the maximum would be, but for us it is not reasonable to get just a few more extra KHS with double or triple the power used. Still feel free to experiment and report your finding and as usual be very careful not to burn your hardware while experimenting with it. If you go with more than 47 kOhm resistor be extra careful if running the device with the bottom of the cooler not attached as you can easily burn the voltage regulators of the ASIC!

Performance Results of the New Gridseed Blade Miner ASIC

5, Apr 2014

gridseed-blade-miner-1

Two days ago we have published information about the new Gridseed Blade Miner and now we already have some performance results available to share with you. Thanks to MinerEU, an official Gridseed EU distributor, we got not only the information about the new device using 80 Gridseed GC3355 chips earlier, but also had the chance to do some performance testing of the new miner and share the results with you. We were provided with remote access to a system that had the new Gridseed Blade Miner attached and we could test with different pools and settings, as well as by using different miner software. We haven’t yet had physical access to these miners, but we plan on ordering one as they are very soon going to start shipping the devices, around 8th or 9th of April or in just a few days. The new Gridseed Blade Miner (80 Chip) is already available for order at a price of $3000 USD at MinerEU.com. But let us get to the tests and results we have performed of the device using remote access…

bfgminer-gridseed-blade-600-mhz-stock

According to the official information from Gridseed one 80-chip Blade Miner is equivalent to 16 Gridseed 5-chip USB miners and it is capable of providing a total hashrate of 5.2 MHS. So we have fired up bfgminer with the Blade Miner configured to run at the stock 600 MHz, the default frequency for the 5-chip miners to see what kind of performance we can get. A bit of a disappointment as we got just about 4 MHs total performance from the two PCBs with 40 chips each that combined make the 80 chip Blade Miner. And since the Blade miner has two PCBs with 40 chips each and different power supply and USB connection in the miner software you see it as two separate devices.

bfgminer-gridseed-blade-700-mhz

If you were hoping that the new Gridseed Blade Miner will be capable of 5.2 MHS as per official specifications at 600 MHz, you will probably not be very happy as it seems that Gridseed has apparently decided to overclock the device and cite specifications at higher clocks. So we went further and ran bfgminer at 700 MHz frequency to see the performance level that we can get and we got a combined hashrate of 4.6-4.7 MHS, so still a bit below the official specifications.

bfgminer-gridseed-blade-800-mhz

Running bfgminer at 800 MHz has allowed us to get us a bit higher than the official specs of 5.2 MHS, we got at up to about 5.6 MHS with these settings with no HW errors to just a few. This means that you can expect to easily run these miners at 800 MHz with no problems and get a bit over the official specs in terms of hashrate. We went down to 750 MHz just to check that with it we are actually getting around 5-5.2 MHS hashrate from the Blade Miner, so that is around the “default” frequency that the new Gridseed Blade miners are supposed to be ran in order for you to get about 5.2 MHs.

gridseed-80-chip-blade-poolside-800-mhz

The image above shows the actual poolside reported hashrate, the first result is from a voltmodded 5-chip Gridseed ASIC running at 490 KHS local hashrate and with 64 worker difficulty and autoswitching to the most profitable coin at scryptguild. The second worker is the new Blade Miner running at 256 worker difficulty and manually selected to mine for LTC with a hashrate of about 5.6 MHS or pretty much the same as the locally reported hashrate. We have found out that setting the worker difficulty to 256 or 512 and not using auto switching does provide better results and less stales (higher efficiency) and you can easily compare the number of stale shares to the 5-chip miner and the 80-chip one and see that for yourself.

bfgminer-gridseed-blade-850-mhz

We also had to try running bfgminer at 850 MHz, a frequency that is the optimal one for Scrypt only mining mode on the 5-chip Gridseed ASIC devices for overclocking without any modifications. Unfortunately the new Gridseed Blade miners do not handle that well the 850 MHz and started producing a bit more than the desirable number of HW errors. As a result of the more HW errors at that frequency the actual performance in terms of hashrate (~5 MHS) decreases as compared to running the miners at 800 MHz, so we do not recommend to go that high with the new ASICs. It is possible that 850 MHz or even a bit more could be achieved after optimizing the cooling of the chips for example, however before actually getting the hardware in our hands we cannot confirm if that is possible or not.

gridseed-80-chip-blade-poolside-838-cgminer

So far we’ve been testing with bfgminer with Gridseed support, the standard miner we used for the smaller 5-chip devices and it works very well with the new miner. It might need a minor tweak to report all local hashrates properly, but the actual performance gets properly reported (the third value) that takes into account the rejects and HW errors. We’ve decided to give a try to cgminer with Gridseed support as well to test with and to find the best frequency in between 800 and 850 MHz that can provide optimum results in terms of hashrate and while minimizing the number of HW errors. We were able to get much better results at 838 MHz as compared to what we were getting at 850 MHz and with some extra hashrate reported poolside. The problem when using cgminer for Gridseed is that the locally reported hashtate is incorrectly reported, though the number of accepted and rejected shares and the number of HW errors seems to be properly reported and poolside performance shows that the miner is properly working and only the reporting is not working right. This should be easily fixable with an update on the software miner code in order to properly report the local hashrate based on the number of chips for each of the PCBs in the Gridseed Blade Miner.

The good news is that you have some headroom for overclocking the new Blade Miner and squeezing some extra performance out of the box over the standard 5.2 MHS. Then there is also the possibility of implementing an easy voltmod in a similar manner as with the smaller 5-chip Gridseed ASIC devices on the bigger Blade Miner and getting some extra boost in performance up to about 8 MHs. Before actually getting the hardware in our hands and trying out we cannot confirm if this is real option or not, but we already suspect that it might be possible. You should not have any problems starting to use the new Gridseed Blade Miner as soon as you receive the device in your hands as it works with the already available software for the smaller miners and if you are familiar with them you should be up and running in no time, especially if you consider the information about the different clocks and hashrate we have provided based on our remote testing. Do note that the Gridseed Blade Miner that we have tested is an earlier unit, so there is a chance that the upcoming units that should start shipping in a few days might be doing even better, though there is no guarantee for that.

We are already remotely testing a second Gridseed Blade Miner ASIC device, again remotely in order to compare two of the devices and see if they would behave differently and is there some variance from unit to unit. We are soon going to post some results from the two Blade Miners and these should also be able to give you a good idea on what hashrate you can expect from two of these ASIC devices, so stay tuned for more information soon.

– For more photos, specifications and information about the new Gridseed Blade Miner…

Another New Voltmod Update for the Gridseed 5-chip BTC/LTC ASIC Miners

1, Apr 2014

gridseed-asic-r52-voltmod

Today we’ve been experimenting with another alternative voltage modification that has been reported by a reader, the previous mods we have done did require solder bridges and replacement of resistors and the new one relies on a replacement of a single resistor. By replacing the R52 resistor (the resistor on the left of the R52 marking, the actual resistor we are replacing is R139) we can increase the voltage and thus be able to overclock even further the Gridseed ASIC devices, so we did experiment a bit with this modification today.

The default resistor is a 402 type 33 kOhm one and the voltage we have measured across it was 1.1925V. The recommendation we got was to to use a 47 kOhm resistor, so we did just that, after replacing the standard 33 kOhm with a 47 kOhm 402 1% we have measured a new voltage of 1.6885V. This is a significant increase in the voltage that we knew would also lead to significant increase in the total power consumption of the device. The chips might be able to handle even higher voltage and allow for more overclock, however the higher the voltage, the higher the chances are that you will shorten significantly the life of the ASIC device. So if you decide to go for a higher voltage the increase in overclock frequency might get lower and lower, but the total power consumption will continue to increase significantly.

gridseed-asic-r52-voltmod-cgminer

After the voltage modification we were able to get stable performance at 1163 MHz overclock with zero to just a few HW errors per hour which is very acceptable for a local hashrate of about 495 KHS. The poolside results we’ve got with this overclock were about 480 KHS, so very good results in terms of performance after doing this single resistor voltmod. Note that you may have varying results on different miners, some might be able to get up to about 1200 MHz, on others you may need to lower to about 1150 MHz. Regardless of that the performance you can expect to get after doing the 47 kOhm resistor mod is close to 500 KHS or about 100% overclock from the stock 600 MHz and with about 65% performance increase. The expected overclock in Scrypt only mode is about 850 MHz and produces about 360 KHS local hashrate with low to no HW errors. Doing the voltmod can give you about 135 KHS more than what you get with the standard overclock, but also rises the question how much it will increase the power usage.

gridseed-asic-r52-voltmod-power-usage

The total power consumption we have measured with a voltmoded Gridseed ASIC running at 1163 MHz was about 20W (including the fan), so it is pretty much double than what we were getting with the previous voltmods. So the big question is it worth to do this extra high voltage modification in order to get a bit higher hashrate? On the short term it is worth it to be able to squeeze every last bit of performance you can get from the ASIC devices you have in order to get a faster return of investment and on time before the big and powerful ASIC devices start shipping later this year. On the long run however you may significantly shorten the life of the device, we cannot currently say how the significant increase of the voltage may affect the normal operation of the device a few months ahead. So a word of caution, be careful and be aware that this mod may significantly shorten the life of your ASIC hardware. Also going for higher voltage by using higher value resistor may just help you get just a bit more extra KHS, but will surely lead to even more significant increase in the power consumption.

We are going to be giving more extensive testing of the 47 kOhm modification in the following days and reporting the results. Most likely we are going to perform the modification to a few more miners in order to see the variance as well. Also going for 20W total power consumption is Ok for us, but 30-40-50-60W+ with minor extra performance on top is not worth it, though if you wish you could try with a 68 kOhm resistor for example for the maximum overclock, but do proceed with extreme caution as it will increase the voltage and power consumption quite a bit further than with 47 kOhm (voltage regulators may die with too much voltage and over 60W of total power consumption!!!). Do note that playing with voltages is dangerous and can easily lead to damaging your hardware. You should also be very careful with the cooling as you increase the voltage, so do monitor carefully the temperature of the GC3355 chips when you have the cooler disassembled and the unit is working in order for example to measure the voltage. And another word of warning, since 402 resistors are pretty small in size and hard to solder for many people, this mod is not suitable for inexperienced users, so better find somebody that can do it for you if you are not sure you will be able to handle it yourself.


top