Sniper G.1 FMA88x / AMD A0-6800K Gigabyte G.1 Sniper A88X Price: $146 on staticice Socket: FM2+ Formfactor: ATX AMD A10-6800K Price: $168 on staticice Socket: FM2+ Frequency: 4.1Ghz - 4.4Ghz (Turbo) Cores: 2 Module / 4 core (Thread) TDP: 100W Unlocked: Yes ------------- Gigabyte launched its Sniper series with the Intel x58 Chipset a couple of years ago now. Since then, the series has trickled down into mainstream sockets, and now, for the first time, onto an AMD platform. They chose none other than the brand new FM2+ platform to do this. FM2+ and A88X : A future proof FM2 For those out of the loop, FM2+ is an evolution of the existing FM2 socket, and provides future support for the upcoming Steamroller/CGN based "Kaveri" APU's. In the meantime though, the Richland based 6xxx series APU's will slot into the backwards compatible socket. And in this case, it's the AMD A10-6800K, to put this combo though its paces. A88X Chipset The flagship chipset for this platform is the A88X. This chipset is an evolution of the A85X, specifically to support pci-e 3.0 for the Kaveri APU, along with other niceties like Configurable TDP (A first on AMD's desktop platform) Everyting else is business as usual, and for existing APU's, it's largely the same. Motherboard Overview In the Box: Let's break this box open! Nothing unusual in the way of accessories, just the essentials here. A (fully labelled) Backplate and Sata cables. For a board of its class, it looks pretty impressive. The colour scheme may not be for everyone, but it is striking, and modern, without much tacky bling to speak of. The Matt black solder mask is classy looking also. What stands out most though is the audio side of things.. which we will get into further on. A gold plated sheild over the Realtec Audio IC, and the 'LED path lighting' when powered up. Layout and VRM The board supports Crossfire in 2x 8 mode, and both of these slots have good spacing, as is expected these days. In between these, a sprinkle of pci 1x slots, and somewhat surprisingly a PCI slot.. Perfect for your 56K modem. .. Oh, and a serial port header! On the back panel we have our gold plated audio jacks, which worth nothing, are a Lotes item, a brand well known for quality sockets. Hidden in between on its side is the gold plated USB DAC port. The onboard video is available via HDMI, DVI and D-SUB, giving support for up to 3 Monitors. One thing I'd like to see now more often is all digital outputs (as one of Gigabytes ITX boards offers) but not to be in this case. Voltage regulation is taken care of by a fairly middle-of-the-road spec 4+2 phase VRM. controlled by an ISL 6377 Smode IC, and using Driver-Mosfet output devices. Given the target position of this board, this is fitting. although it will limit those wanting to go all out with high voltage OC's.. if you're content with a mild OC, or of course you're going to the Underclocking / Undervolting route, then well it's perfect. All about the Audio Now without a doubt, the most stand out thing of this board is the focus on audio. They really have gone to town here in comparison to your ordinary implementation. Now given its common for companies so list a lot of fluff that doesn't man anything. Let's look at exactly what they've done that at least has potential to make a difference.. For the onboard audio: * Socketed, upgradable output stage with Buur brown OPA 2132OP-AMP *2.5 /6 gain selection * Shielded Audio codec IC * Nichlicon audio grade electrolytic capacitors * Physical segrgation of Audio circuirty * Gold plated Audio jacks Seperate to the onboard audio, it Also features Gigabytes "USB-DAC " Port, which is designed to be used for USB powered DAC's. The idea is by having it's own little swithcmode power supply, it is immune to load- induced noise on the 'normal' 5v line used by your ordinary USB ports.. You can see the component of this cute little supply here. Further on we will look at real world performance of this in more detail. BIOS Incorporating Gigabytes UEFI GUI, the BIOS is fairly basic on this board.. Which probably won't impress the enthusiast trying to tweak to the n'th degree, but it does have all the essentials. . Click to view full size! Click to view full size! Click to view full size! Click to view full size! Click to view full size! Some items of note. Voltage adjustment granularity is a reasonable 0.006v , also I did notice some translation issues. CPU, CPU-NB, RAM multiplier, base clock, all here. RAM multi maxes out at 2400. Given Kaveri may support this speed, I do wonder if higher multiplier (OC) will become available. APU Overview To put this platform through its paces, we have AMD's latest available APU, the Richland based A10-6800K. I did a review earlier in the year with its predecessor the 5800K. So it will be interesting to at least loosly compare the potential of this refreshed APU Frequency up. Voltage Down Loading up AMDMsrTweaker revealed the VID's and turbo frequencies at various load P states for this particular CPU as P3 4.1Ghz 1.31v P2 4.2Ghz 1.35v P1 4.3Ghz 1.40v P0 4.4Ghz 1.435v So at it's base clock, Vcore will be 1.31v. TEST SETUP Motherboard: Gigabyte G.1 Sniper FMA88X CPU: AMD A10-6800K RAM: DDR3-2400 @ 2133-11-10-10-10 (stock) 2400 (OC) HDD: Samsung 840 Evo 256GB As per my last Review, I'm not going to run through 50 benchmarks @ stock speeds. There are an equal number of commerical review sites you can pour over for this. One thing you won't find in existing benchmarks though is the new Cinebench R15. Hence I've included it here.. at stock clocks the 6800K churns out a score of 328cb / 101cb Single core. OCing - stock Heatsink.. As has been commented by many before. The stock heatsink, while "adequte" really isn't much use for overclocking. As seen above, at 4.1Ghz, the VID is 1.3v. Disabling Turbo, and Setting this voltage really didn't result in anything worthwhile. I was only able to reach 4.3Ghz reliably at this voltage. So there's not a lot of wiggle room at lower vcore. This is not entirely unexpected. Richland has brought with it much finer control over frequency and voltage P states, and also a more mature process = stock frequencies a lot closer to the ceiling... Bumping Vcore up a notch to 1.35v set, (but still well below Turbo VID) yieled 4.4ghz stable. With Load temperatures already approaching 80c with the stock heatsink.. Time to change it out. Overcclocking - Aftermarket Heatsink. One of the great things about AMD's sockets, is their mounting system hasn't changed in the better half of a decade now. Whilst Intel's basic design hasn't either, they keep changing the mounting hole centers.. which = useless, but still working heatsinks everywhere. There are some incompatabilties with 939 heatsinks moving to AM2. But from that point on, everything has been compatible. So With my Noctua busy.. I bring you , OId faithful. Orginally bought back in 2007 , it was fairly popular bargain then, . with 4 silicon rubber mounts, around the whisper quiet, high AOA 92mm fan. The moral? Well, given the limitations of the stock HS, it means one don't necessarily need to go spend money on quiet heat sink's.. If you've had AMD systems before, you might just have something like this hidden in that storage box.. Whilst there has been a lot of progress in HS design. They're still a lot better than what comes out of the box. Strapping old faithful on, with the existing settings, Load temp dropped by ~12c immediatly. With some thermal headroom to play with, and the vcore bumped up further at 1.47v, I managed a stable 4.7Ghz. I was able to bench at 4.8Ghz with no troubles, but long term stabilty was an issue as heat set in.. With a better heatsink, There's no reason to doubt this would be an achiveable 24/7 clock. With this board not really being aimed at extreme OC's, this is quite respectable.. and an improvement on Trinity, despite a lower voltage: RAM Overclocking With Richland Supporting DDR3-2133 out of the box, and the Board's Multi maxing out at DDR3-2400, I wasn't left with much to tweak here. I had some issues trying to manually set 2400, so I set the XMP profile for 2400 and bingo, it booted up and passed memtest without hassle. This profile sets a DRAM voltage of 1.65v, which is the typical rated voltage for 2400 Kits, though you could probably play with this and bring it down. <Overclocked Performance> Around 13% increase in performance on these couple of benches. roughly inline with the fact APM was Turboing cores to 4.2 even 4.3Ghz occasionaly at stock. With No Turbo mode after overclocking, single threaded performance see's even less uplift here.. AMDmsrTweaker - OC Turbo Profiles As such, something I'm always keen to play with is utilizing turbo mode whilst overclocking. This has proved difficult to do sometimes when using the BIOS options. So One of the most reliable ways to do this currently, is a little application called AMDMSRTweaker. This command line program is the Sucessor to the GUI based PhenomMSRTweaker, and like it, allows you to set your P states using a simple command-line syntax. By making simple batch files up you can apply an OC ( or UC) "Profile" In seconds. To experiment with the above 4.8Ghz OC, I set some turbo P state's to 4.9 and 5.0Ghz. One of the issues this presents, is the well known "Throttling" Effect caused by the Application Power managment. this is what controls turbo functionality, using current and temperature sensing, and tries to keep power consumption in check.. This is also how Configurable TDP works, when it's available (not here) When Overclocked, and Overvolted though, behaviour becomes a bit unpredictable.. Surprisingly, it does not seem to be too fussed about keeping TDP completely within the 100w limit. Whilst it does throttle, it didn't seem to ever drop below 4.7ghz, though it is hard to tell what it is doing in 'real time'. And with that, The Cinebench result below seems to reflect a 4.8Ghz average frequency.. at 366cb. So clearly, even though clockspeed may not be 100% predictable / "set in stone" , it still does increase performance pretty reliably. Running the Single core benchmark though, had it pinned at 5 Ghz, and here we see a small but definite advantage in single threaded performance.. At the end of the day, as we're approaching the limits of clockspeed on air, even at high vcore (5ghz) there's not much point in this.. At lower clockspeeds, and when Underclocking however it certainly becomes worthwhile. Cinebench R15 - Fixed Overclock Vs Turbo Overclock I would really like to spend a lot more time experimenting with this, perhaps another day. Once confingurable TDP is something easily available to Enthusiasts it should be better to play with Power Consumption and Underclocking / Undervolting[/CENTER] Because of the absense of in BIOS options for undervolting, I had to use AMDmsrtweaker exlusivly for these tests.. This little program is so good though that's no big deal at all. Using my pico=psu and Prova Dual channel Multimete,r I was able to aquire DC power intput into the pico-psu. This means, near 100% efficency on the 12v supply (the important one) , and ~90% on the other rails. Giving much more accurate platform power consumption readings than a normal 240v input measuremnt. Because of Richlands advanced power managment, we see extremely consistent peak power consumption. After a short burst of up to 120w or so, the Power managment would ensure the platform would not exceed 110 Watts, regardless of type of program. To achieve this we do see some throttling.. At various intervals during a benchmark run, it could be seen to jump down to the 3.8Ghz P3 state for a fraction of a second. For the most part though it was turboing to at least 4.2Ghz Stock clocks, Undervolted Power profiles. After a bit of trial and error, I settled on approximatly -0.075v on all power states. Anything less would bring instability. Clearly, with Richland, as we saw with Overclocking tests above, there's not a great deal of wiggle room. Underclocking. Drop clockspeeds a little though, and things start to look up.. or should I say down in regards to power consumption. dropping the P states to 3.8 - 4.1Ghz yields a big drop in load consumption.. But. Because consumption has dropped, APM keeps clock frequencies up around 3.9-4Ghz right through a cinebench run. The net result is less than 10% drop in Cinebench score. By the time we've underclocked to 3.2/3.6Ghz Turbo, Load consumption is only 58w at the wall. A 46% increase in perf/watt. Also worth a mention is single core power consumption, which also dropped, but surprisingly not that much. Unfortinatly I did not have time to play around with this further, but it's clear, yet again a small sacrifice in performance yields big drop's in power consumption.. and more importantly Performance/watt.. Graphics Overclocking and Performance I spent a bit of time with in my last look at the Trinity system. With richland having the same GPU spec, with a small bump in frequency I was expecting pretty similar results, and that's what I got. Overclocking was again a breeze, with plenty of headroom. Whilst keeping a Modest +0.1v CPU-NB voltage bump, I hit similar frequencies for Richland. CPUNB: +0.1v GPU Frequency: 1140Mhz Battlefield 4 Beta - Too early to tell, but just playable. Despite the Beta's horribly unoptimised condition, armed with the Overclocked settings above, I did manage to squeeze somewhat playable performance out of the GPU in this upcoming title on its lowest settings @ 1080p Bf3 Performance, as expected mirrored my testing with it's predecessor.. Running amost identical clockspeeds throughout, just newer drivers. At low settings FPS hover around a pretty 'playable' high 30s to 40 FPS. The Crucial part here of course is RAM speed more so than GPU clock. In game Settings (BF3 & 4): Resolution: 1920x1080 IQ Settings: LOW preset / No AA/AF Onboard audio - no ordinary onboard.. As I said earlier, Gigabyte have put quite a focus on improving [primarily 2 channel] audio on this board. Both the onboard Realtek audio, and with the provision of the USB DAC port We've all heard the arguments against onboard sound.. Often pretty once sided against!. I've even entered the odd one myself. Usually proclaiming that it's not as bad as people make out, at least when driving an amplifier.. One thing they're certainly not good at though, is driving headphones, and having a low noise floor. Utilizing an OP2134 BB opamp, the output stage is configurable with a 2.5x gain , which with independant L/R switches can be upped to 6x for high impedance headphones. Now it's unlikely anyone would be swayed from a dedicated soundcard + Amp to try this. But if you don't have these things already, and run headphones. This is a board for you.. Testing.. To be continued. I spent a good part of an evening trying to get my old EM-U 404 soundcard working, but alas, it was not to be. This used to be my primary card, and one I used for testing distortion etc, due to it's very high quality (at the time) inputs. Unfortunatly she is dead. May it RIP. So, testing will have to wait for the moment. Subjectivly though, well, it's very obvious the added gain packs a punch when driving headphones. Whilst it seems there is some rolloff with bass at this higher gain, possibly caused by the current capability of the Op-amp. This will definitely be something worth testing when I can. Dedicated USB Port- Worth it?.. Past experience has told me that one of the biggest causes of noise in both onboard audio, and USB supplies is caused by high current switching from things such as gfx cards. Gigabytes decision to include a USB port with it's own s-mode regulator certainly got my interest. So To test Gigabytes claims out of lower noise, I gave it it's best chance to prove itself, by fitting a discrete GFX card (HD5850) ,something I know causes a noisy supply from experience, and doing a run of 3dMark firestrike. Both at idle, and during the 3dmark run, I Measured the noise level of the 5v USB supply on both standard USB port, and the USB_DAC Port To make the results relevent, I set a 140Khz Filter, to focus on noise that will effect the Audio spectrum. Standard Port Idle 3dmark run HDD transfer USB DAC Port Idle 3dmark run Well both my ears, and the results speak for themselves here. The noise level in this band is markedly reduced. during 3dmark, The standard USB ports had a 130mv PP ripple, in tune with the FPS of the video card, thus, the current pulses on the 5v line as the card renders each frame. The powered Port however, only a very small ripple has penetrated it's Smode Supply. Just as noticable on the waveform is the spurious noise even with the machine sitting idle, likely caused by disc activity Clearly If you're using a USB power DAC, this is the port to plug it into Conclusion. When I first looked at the new AM2 Platform early this year, I came away pretty impressed overall. Whilst Performance/watt was still far behind Intel on the CPU side , it was made up for by class leading graphics, low idle power, and of course Overclockabilty.. Something not available in this price range/ class from Intel's Offerings. So What's changed? . Well, with the Release of Richland, CPU clockspeeds have gone up, Core voltges gone slightly down. and at the same time, they're also undoubtedly more overclockable than the previous Trinity CPU's. This has at least allowed the platform to remain in a similar position after Haswell's release, as it was bbefore when going head-head with Ivy based i3s.. Whilst the Haswell is a NEW Architecture, The net result is the same.. i.e an incremental performance increase. Without having 65w APU's to test, it's hard to say how far behind Intel perf/watt remains, but repeating The underclocking exersize on the K series APU showed big gains to be had here. As for A88x.. Well, with current APU's there isn't much to speak of. The big thing is, future Kaveri APU Support. So at this point, if you're looking an an APU build. an FM2+ Platform is the obvious choice, even if it doesn't offer any real world advantages at this point. Now to the G.1 Sniper.. Given the unique features of this Board, it's a little hard to make a generalised conclusion.. but on the whole, even without the impressive audio features, this is a pretty decent board for the money. It did pretty well on the bench. Overclocking was quite painless, like the higher end A85X board I looked at last time. There was no funny business with the BIOS, no coil whine, and remained stable. Perhaps the only disapointing part for the enthusiast, (at least pending any updates) is the very basic BIOS options, and as was discovered, no Undervolting options - simply offset type Voltage adjustment (up only). So if you're a heavy BIOS tweaker, well it's probably not going to excite you in this department. Being a passion of mine, The effort gone into audio really did impress me though. Bar the lighting effects.. Just about none of the extra features could be called a "Gimmick" Whilst I don't any data to suggest the effort put into shielding, and isolating the audio IC has effects on it's output (yet - and I refuse to do subjective tests on this!), the adjustable gain and Opamp are good features regardless, and really make driving headphones directly from onboard 'doable' As I'm sure I've said before.. I would really like to see these things filter these features down in to mATx and mITX form factors! Even if means only supporting 65w APUs, that's fine. Kaveri is rumoured to be (at least for the majority of SKU's), a 65w TDP anyway. Big Thanks to AMD and gigabyte Australia for the review samples!.. Over and out.. Any questions or requests for more/ better info, don't hesitate to ask, I will do my best !