Overclock.net banner

1301 - 1313 of 1313 Posts

·
Registered
Joined
·
477 Posts
Fantastic Job! Great to see you have finally finished it.<br><br>
Though those cables could do with a bit of tidying, how about something to conceal them like a chrome exhaust? (just kidding <img alt="tongue.gif" class="bbcode_smiley" src="http://files.overclock.net/images/smilies/tongue.gif"> )
 

·
Chief Over-Engineer
Joined
·
608 Posts
Discussion Starter #1,303
A little addendum to the 2nd Presentation Picture set.<br><br>
With a bit of effort I have managed to get a couple of vaguely decent photos of the two-tone reservoir lighting, which looks fantastic but didn’t show up well in the selection posted previously. There are also two animated .gif files showing that effect and the RayStorm heartbeat program.<br><br>
But first, once again, and as ever, my thanks to you all for your very kind comments. It must be stated for the record how genuinely overjoyed, and frankly rather humbled, I am that you all like <i>Ironbeast</i> so much - and that you’ve actually enjoyed following my snail’s-pace build log, for however long or not that may have been, even though it didn’t quite keep up the quality of writing that it started with. You should know that this thing would never have been finished, maybe never even clawed its way out of the CAD files, if it were not for your support and enthusiasm for the idea.<br><br><div class="bbcode_center" style="text-align:center;"><i>Ironbeast</i> is as much your build as mine.</div>
<br>
And to those who were here at the beginning, or the middle, or even back when I first fired this thing up, and who are still reading this now – how on earth have you put up with me for so long?<br><br>
For the absolute die-hards there will be one final formal log entry – <i>Ironbeast: Project Evaluation</i> – ready for you shortly; a one-post overview to set out which of the ideas worked and which didn’t, to give some of my thoughts and reflections on the build, and generally wrap this up. You can expect that just as soon as I finish writing it...<br><br><div class="bbcode_center" style="text-align:center;"><br><a class="H-lightbox-open" href="http://www.overclock.net/content/type/61/id/3096929/"><img alt="" class="lightbox-enabled" data-id="3096929" data-type="61" src="http://www.overclock.net/content/type/61/id/3096929/width/600/height/1200/flags/LL" style="; width: 600px; height: 542px"></a><br><br><a class="H-lightbox-open" href="http://www.overclock.net/content/type/61/id/3096936/"><img alt="" class="lightbox-enabled" data-id="3096936" data-type="61" src="http://www.overclock.net/content/type/61/id/3096936/width/600/height/1200/flags/LL" style="; width: 600px; height: 546px"></a><br><br><a class="H-lightbox-open" href="http://www.overclock.net/content/type/61/id/3096935/"><img alt="" class="lightbox-enabled" data-id="3096935" data-type="61" src="http://www.overclock.net/content/type/61/id/3096935/width/500/height/1000/flags/LL" style="; width: 352px; height: 288px"></a><br><br><a class="H-lightbox-open" href="http://www.overclock.net/content/type/61/id/3096933/"><img alt="" class="lightbox-enabled" data-id="3096933" data-type="61" src="http://www.overclock.net/content/type/61/id/3096933/width/750/height/1500/flags/LL" style="; width: 750px; height: 422px"></a><br></div>
<br><div class="quote-container" data-huddler-embed="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26274602" data-huddler-embed-placeholder="false"><span>Quote:</span>
<div class="quote-block">Originally Posted by <strong>donkidonki</strong> <a href="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26274602"><img alt="View Post" class="inlineimg" src="/img/forum/go_quote.gif"></a><br><br>
Fantastic Job! Great to see you have finally finished it.<br><br>
Though those cables could do with a bit of tidying, how about something to conceal them like a chrome exhaust? (just kidding <img alt="tongue.gif" class="bbcode_smiley" src="http://files.overclock.net/images/smilies/tongue.gif"> )</div>
</div>
<br>
A <i>bit</i> of tidying? Oh you’ve no idea… the cabling at the back is an unbelievable mess. As determined as I am to say that this is as far as the project goes, I've no doubt something will have to be done about that – though it will probably have to wait until <i>Ironbeast</i>’s upcoming clean and coolant change.<br><br><div class="quote-container" data-huddler-embed="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1290#post_26271246" data-huddler-embed-placeholder="false"><span>Quote:</span>
<div class="quote-block">Originally Posted by <strong>Angry-Hermit</strong> <a href="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1290#post_26271246"><img alt="View Post" class="inlineimg" src="/img/forum/go_quote.gif"></a><br><br>
So whats next <img alt="biggrin.gif" class="bbcode_smiley" src="http://files.overclock.net/images/smilies/biggrin.gif"></div>
</div>
<br>
Next up is all the ‘Stage 2’ stuff that I’ve occasionally referred to. Buying new peripherals for this thing has been consistently put off until <i>Ironbeast</i> was in a finished state, such that most of it is now very outdated <i>(monitor, keyboard)</i> and/or falling to bits <i>(mouse, headphones)</i>. There is also the half-finished CAD model for a replacement desk design that was intended to house it all, and to be the lair of the beast itself. This needs some more work to complete, then fabrication drawings for the parts, and then it needs building. After, or potentially overlapping with that, would be one of three potential new projects that I have some early concept work for; designated at present as <i>Precursor</i>, <i>Timberwolf</i> and <i>Ironbeast Mark II</i>.
 

·
Registered
Joined
·
293 Posts
Awesome, I hope to start on a desk build soon, more of a prototype for a sit to standing mechanism, so when I actually move to a house Ill know what not to do when I build a final one. Make sure you update this thread if you start something else. Be glad to follow it as well.
 

·
Chief Over-Engineer
Joined
·
608 Posts
Discussion Starter #1,305
<p style="text-align:center;"><em>So we come now to the very end. This will be the last of my regular entries for this build log - a little conclusion (i.e. a big rambling wall of text) just to wrap everything up. Any questions and comments are still more than welcome (even if you’re reading this sentence years after I typed it) and there may well be the odd addendum in future if anything particularly noteworthy happens. Otherwise this brings The Ironbeast Project to a close - for the Mark 1 at least</em>…   </p>
<p> </p>
<p> </p>
<p> </p>
<p style="text-align:center;"><span style="font-size:24px;">IRONBEAST</span></p>
<p style="text-align:center;"><span style="font-size:48px;">Project Evaluation</span></p>
<p style="text-align:center;"> </p>
<p style="text-align:center;">21<sup>st</sup> August 2017</p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"><em>A retrospective on this very lengthy build, outlining its various successes and failures.</em></p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"><em>(This post really is only for the staunchest die-hards, and probably takes the prize for the longest of my log entries. Bullet point summaries have been included to try to help with that, and I have done my best to stay suitably objective and avoid letting this degenerate into an inglorious lament bemoaning of my all my self-perceived failures…)</em></p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"> </p>
<a class="spoiler-link H-spoiler-toggle" href="#"><strong>Reflections on the Concept: The experimental ideas, and things which didn't get done.</strong> <span class="spoiler-help">(Click to show)</span></a><div class="spoiler-hidden">
<p style="text-align:center;"> </p>
<p style="text-align:center;"><span style="font-size:18px;"><span style="text-decoration:underline;">Concept</span></span></p>
<p> </p>
<p>As I hope is evident to everyone seeing my final photographs, and as was fairly clear to those who first saw my initial renders, <em>Ironbeast</em> is a markedly unconventional machine in a lot of ways. The earliest concepts <em>(the proto-Ironbeast, long before this build log started)</em> was far less ambitious, much more in line with the many other fantastic Corsair 900D-based builds that originally inspired it. It is very much the untested bits, all the ways that this build diverged from the normal approach, which has made it into what it is – that unique “something special” that I so wanted it to be.</p>
<p> </p>
<p>To summarise, the main experimental features of the <em>Ironbeast</em> design were:</p>
<p> </p>
<ul><li>The symmetrical internal layout.</li>
<li>The recessed and centred motherboard position.</li>
<li>The concealed & tightly spaced offside pipes and wiring.</li>
<li>Pipework sections constructed entirely from fittings.</li>
<li>A smooth 4-way GPU flow split & return with manifolds.</li>
<li>Crosslinked loop configuration.</li>
<li>Clear frameless front panel.</li>
<li>A colour scheme defined only by the lighting, fully adjustable according to mood.</li>
<li>Routing the light from source LED strips using side-glow optical fibre.</li>
<li>Individual control of each light source via Adafruit Neopixel addressable RGB LEDs.</li>
</ul><p> </p>
<p>Considering that I had absolutely no idea whether any of this was going to work, it is really rather pleasing to be able to say that these concepts have in fact turned out successfully. The construction process has revealed many ways in which they could be improved, or inspired ideas for alternative ways of achieving the same result, but nonetheless they all work. In many cases the retail market has since developed similar designs, so many of the things that on <em>Ironbeast</em> had to be created bespoke are now much easier to achieve - or indeed possible for you to buy straight off the shelf.</p>
<p> </p>
<p><strong>Omitted & Unfinished Elements</strong></p>
<p> </p>
<p>Though a lot of what was planned for <em>Ironbeast</em> has, at long last, come to fruition <em>(in one form or another)</em> this was nonetheless an extremely ambitious <em>(preposterously foolhardy)</em> design for a first build. As such there are more than a few elements which have had to be scrapped at various points along the way - or which still remained incomplete as of a month or so ago when I made the decision to suspended any further work and wrap this project up. These include:</p>
<p> </p>
<ul><li>My own custom designed side-port GPU waterblocks. <em>(One of the first things to go, and in truth always acknowledged as an unrealistic proposition - though the concept work was included earlier in the build log.)</em></li>
<li>Custom X99-E WS full-cover waterblock. <em>(Rendered unnecessary. Bitspower eventually came to my rescue with their own design, something I long thought no manufacturer would ever find it financially viable to market)</em></li>
<li>Comprehensive cooling loop monitoring system. <em>(Eliminated due to restricted wiring space and as the MPS sensors would not quite fit. Replaced by a much more basic affair. Rather sad about that one if I’m honest...)</em></li>
<li>Roller base. <em>(Redesigned in favour of using Saturn-type bearings instead).</em></li>
<li>Supplementary/alternative external radiator array, chilled-air booster box. <em>(Never developed beyond the basic concept, though the required expansion ports in the loop do exist to support it.)</em></li>
<li>Motherboard cover plate, GPU cover plate, custom SLI-bridge & ‘power bridge’. <em>(Intended to hide the rear I/O and all the power cables from nearside view, thus completing the symmetrical look. Unfinished element.)</em></li>
<li>Replacement of the remaining visible soft tubing with rigid acrylic. <em>(Unfinished element.)</em></li>
<li>Dominator Platinum Lightbars <em>(That upgrade was not a spending priority at any stage, and so never happened.)</em></li>
<li>Front panel edge lighting<em>. (Concept successfully tested, but never implemented.)</em></li>
<li>Rear panel cable tidy. <em>(Never completed, but it may well prove ultimately necessary to do so.)</em></li>
<li>Dedication plate, logo on top GPU backplate. <em>(May still be done - call it sentiment but I don’t want to omit this.)</em></li>
</ul><p> </p>
<p>Beyond these things there also hasn’t been a proper performance evaluation. Construction has taken so long that there has never been sufficient free time to do any of the proper flow and thermodynamic stress tests that were promised <em>(to let us find out how well Ironbeast actually functions - as a computer, rather than an art installation...)</em> It would still be possible do them, there’s nothing specific stopping me, but the decision was made to end things here and I feel I must hold to that - otherwise a new project will never get started. Judging from everyday experience, the beast works – indeed it works very well - and a few basic real-world temperature tests were included earlier in the build log. This however is not the scientifically rigorous empirical performance testing which I hoped to do, something which would have been both intellectually interesting and usefully informative for a novice builder like me. Likewise there hasn’t been any proper methodical overclocking of the system to see how much performance it truly has <em>(it’s tuned up a bit, but not that much)</em>, which is rather sad since it would have been nice to have tried for a respectable 5GHz+ spot on the Haswell-E leaderboard, even this late in the day. <em>(One of the dreams - I guess that’s the price for being so slow…)</em></p>
</div>
<a class="spoiler-link H-spoiler-toggle" href="#"><strong>Reflections on Technical Aspects: Case, System Hardware and the Cooling Loop</strong> <span class="spoiler-help">(Click to show)</span></a><div class="spoiler-hidden">
<p style="text-align:center;"> </p>
<p style="text-align:center;"><span style="font-size:18px;"><span style="text-decoration:underline;">Technical Aspects</span></span></p>
<p> </p>
<p><strong>The 900D</strong></p>
<p> </p>
<p>Self-evidently missing from the above list is the elephant in the room, the <em>Corsair Obsidian 900D</em> case on which <em>Ironbeast</em> was based. When I finally committed to doing my proposed self-build computer in autumn 2014, the 900D - which I had selected as my choice some months before - was bought as its chassis. <em>(By this time the design sketches were recognisably Ironbeast, but still some months away from the fully rendered CAD model with which this build log started)</em>. The option to continue with an off-the-shelf case was pursued largely because I thought myself lacking, in both time and talent, to attempt a full scratch build. <em>(Yes I know, hindsight…)</em>. For those that were not following back then, the intention was to remove the front drive bays – as inspired by <em>Jameswalt1</em>’s epic “Robocop” – and to replace the motherboard tray with a custom one to support the centred and recessed layout. Then of course my overexploited and downtrodden old Alienware waved the white flag, leaving me PC-less and with a stark choice: back out, either buying off-the-shelf or doing something conventional, or commit to building a design which - by my standards at least - wasn’t yet ready.</p>
<p> </p>
<p>When we started things here the plan was still this conversion of the 900D. Of course, as some of you will know, a little later I got seriously cold feet about the whole thing – whether parts would line up, whether the manifolds would work, whether CPU-side flow path was too complex and restrictive, etc. etc. As such I decided to build a timber-framed mock-up, the <em>Phase 3 Test Frame</em>, a full-size proof-of-concept to reassure myself the layout design was viable. <em>(The Phase 1 being my autumn 2014 scrap-timber test bench, to jury-rig the hardware into an operational state, and Phase 2 was the later quad-GPU version of it)</em>. This decision was quite fortuitous in the end, as getting any of this kind of metalwork done proved impossible – contrary to assurances I’d been given before starting – and the 900D conversion had to be scrapped. And yet, thanks wholly to support from you lot, we carried on and attempted to make something broadly similar using the test frame instead. With the benefit of endless little tweaks and revisions, and underneath all its spray paint, wrap, trim & custom case panels, that one-time proof-of-concept mock-up <strong><em>is</em></strong> the <em>Ironbeast</em> that you see today.</p>
<p> </p>
<p>Now if I’m trying to be objective, it would be fair to say that the build has turned out comparatively well for a first go - especially given the ambitious nature of the original proposal, and then the complexities of modifying the test frame to passably resemble it. <em>(Phase 3 was only ever intended to test the layout – it was never designed to actually <strong>look</strong> like Ironbeast. Plus, it’s <strong>Iron</strong>beast – the clue is in the name. Going from a steel structure to wood; endless little problems.)</em> That being said, there are just too many design compromises, too many ‘improvised’ solutions, and too many ‘rough edges’ for me to be truly happy with it; even considering that this should - technically speaking - be classed as an operational concept prototype, rather than the actual <em>Mk.1 Ironbeast</em> that was once envisioned. In retrospect the decision to use, and then continue to use, an off-the-shelf case was the wrong one. There are simply too many limitations in trying to do something genuinely unconventional when you are constrained to working within a pre-defined enclosure. In particular the <em>Ironbeast</em> design has major limitations for easy accessibility & maintenance, suboptimal airflow, highly problematic space restrictions <em>(especially for cables)</em> and other unavoidable layout constraints <em>(e.g. the PSU position is deeply annoying)</em>. With hindsight I should have shelved this concept, stuffed the hardware into the 900D using a conventional layout, and then started from scratch on a fully custom design. There would be a better, more modern, and far more practical machine here now - and it probably would not have taken any longer overall than this has.  </p>
<p> </p>
<p>If you think that is all a bit downcast of me, apologies, but it’s in my training <em>(and doubtless in my nature)</em> to be depressingly objective about these things. However, I shall temper it with this more optimistic thought. Looking back on the project and at the completed rig as-is, I am confident enough to say that - had there been proper metalworking facilities available for me to make it, and although it would have been murderous aggro to assemble - the proposed conversion of the 900D <strong><em>was</em></strong> viable, and that for all its flaws the original <em>Mk.1</em> design would indeed have worked. </p>
<p> </p>
<p><strong><em>Key Lessons: Chassis</em></strong></p>
<p> </p>
<ul><li>Off the shelf cases are convenient but limiting.</li>
<li>Bespoke designs are expensive, slow to get made, and require me to spend many weeks lost in CAD.</li>
<li>Getting any custom metalwork made seems to be very difficult and very expensive around here.</li>
<li>Keep sheet-material components to the most basic shapes as far as possible, minimising interior cutouts.</li>
<li>Trespa is an excellent material for thin sheet work and very robust, but it is quite expensive.</li>
<li>Untreated smooth planed ash gives a fantastic look - use it in future builds.</li>
<li>Slightly oversize mounting holes or create slots to facilitate easier positioning, especially of radiators.</li>
<li>DO NOT underestimate the amount of space required for cables.</li>
<li>The recessed and centred motherboard idea works very well.</li>
<li>Far more focus should be directed towards improving ease of construction & maintenance in future builds.  </li>
</ul><p> </p>
<p><strong>Hardware</strong></p>
<p> </p>
<p>On reflection I think the choices for the main system components were broadly the right ones, all things considered. Of course the main lesson - by far - has been about <em>when</em> to buy the technology: get it as late as possible - <em>everything</em> will take longer than you think! In my defence I didn’t have much choice about it this time, but it was still a mistake in a project like this to buy the system backbone before the case and cooling loop were ready - it all just becomes obsolete far too fast, especially graphics cards. Then again, if I hadn’t bought it all, there would be an off-the-shelf PC sitting here <em>(or at best a much more conventional rig)</em> and <em>Ironbeast</em> would likely never have been completed in CAD, much less exist in the real world: yet another half-finished idea gathering dust in a forgotten corner of my hard drive…</p>
<p> </p>
<p>The <em>Intel 5960X</em> was an expensive but good selection, and going for core count rather than potential overclocking speed was the correct approach for my usage. Whilst extra power is always desirable – especially in regards to the CAD work – my Haswell-E has never yet seemed lacking to me. Although the flagship processors of newer series have outperformed it, the cost-benefit has never been enough to make me realistically deem an upgrade to be worthwhile. <em>(I know a lot of people consider Intel’s recent enthusiast grade CPUs to have had lacklustre performance increases compared to previous generations, though still being a relative newcomer – weird as that sounds – I’m not really in a position to comment on that with any authority).</em> It may be that the i9 <em>X-series</em> will be more enticing, or perhaps not. It may be that AMD processors become a viable contender for the heart of a new system. Rest assured that your opinions on the matter will definitely be sought for whatever subsequent project there may be. <em>(Which won’t prompt any arguments at all obviously…)</em></p>
<p>                                                                                                       </p>
<p>The 32GB of <em>Dominator Platinum</em> RAM has been more than sufficient memory for my needs; though with hindsight the merits of getting the higher speed models was doubtful, given that <em>Ironbeast</em> hasn’t been pushed to the ragged edge of performance. It remains my favourite aesthetically, most of the ‘style’ models of RAM are too showy for my taste, but on balance it would have been better to upgrade them with the <em>Lightbars</em> – that is an omission I regret. Whether there were, and are, superior alternatives is also something to be debated in as part of any successor projects.</p>
<p> </p>
<p>At the time of its debut the <em>ASUS X99-E WS</em> was the only <em>X99</em>-based motherboard I really liked - it still is. Though the first one I got was sadly defective, causing me to panic and cast grave doubts on the whole idea of ever having started a custom build, the replacement has served me dutifully and well ever since. <em>(For those that don’t know, a broken WS was not at all uncommon at that time - a lot of people got defective units especially in the first production run. However I’m told my experience of absolutely faultless customer service from ASUS must have been an insane fluke, as it is something that caused more than a few to give up on the X-99E completely).</em> In a similar vein I can offer absolutely no complaints for the <em>Samsung 850 Pro</em> SSDs, or for the <em>XP941</em> M.2 SSD, or for the <em>Corsair AX1500i</em> PSU <em>(which, though really big, has dependably offered all the power I could need).</em> The <em>Aqua Computer Aquaero</em> fan controller is, exactly as you told me, a very good piece of kit for a comprehensive monitoring setup if you know what you’re doing. Though it ultimately proved rather unnecessary in this build as things turned out, I would definitely use it again.</p>
<p> </p>
<p>Graphics was always a difficult choice in this build, as a few of you may well remember. When first envisioned <em>Ironbeast</em> would have had four of the <em>Kepler</em>-based <em>Superclocked GTX Titan Blacks</em> - on account of their ability to do high-end gaming and do FP64 work tolerably well at a decent price<em>. (As I said at the time, double precision compute ability mattered a lot to me back then; but Quadro was financially unviable and unjustifiable for the level of CAD stuff I do)</em>. I will never know if they would have been better - you convinced me to wait for the <em>900-series,</em> the <em>Titan X</em>, then AMD’s <em>Fury X</em> and then the <em>980Ti</em>; at which point Titan Blacks were proving very difficult to get hold of. From this emerged the unusual 3+1 configuration that <em>Ironbeast</em> now has; one Black and 980Ti in 3-way SLI – the former for general purpose use, CAD modelling and PhysX, switching to the latter configuration for gaming, and using all four for multi-GPU rendering work. Now of course the whole thing was made entirely academic by the replacement of my old CAD software suite <em>(and the fact that GPU PhysX is pretty redundant)</em>. However, quite by accident, the 3+1 idea has worked out extremely well for an entirely different reason. The ability to switch outputs between a single card & SLI with a single button <em>(via an HDMI switch</em>, <em>which is slightly temperamental, I think due to my outdated monitor)</em> lets me dramatically reduce <em>Ironbeast</em>’s power consumption whenever 3-way SLI is not required - which is great for the electricity bills.</p>
<p> </p>
<p>In retrospect, the matter of whether four Titan Blacks, Titan-Xs, the AMD Fury, or even some combination of GeForce & Quadro <em>(or Radeon & FirePro)</em> would have made for a better build is a question which will simply have to be left hanging. <em>(Though it must be said that Fury X, being DVI-less, would in hindsight have made for a more convenient choice physically)</em>. However there is one thing to be said for <em>Ironbeast</em>’s graphics setup; the 980Ti predates the death of 3-way & 4-way SLI <em>(and the slow death of SLI in general).</em> I’m rather proud that this machine can rank amongst the last of the breed of monster multi-GPU gaming rigs – our dinosaurs in the age of smartphones – at least for now. For myself I have been waiting years now, through several generations of graphics cards, to see SLI die and a new solution rise <em>(preferably in time for Mk.2)</em>. Maybe it never will. Maybe the limits of silicon will see a big move toward parallel computing in domestic machines. Maybe Volta will bring NVLink to some future GTX Titan card next year and drain all our wallets again. Who knows? Regardless I still hold out hope that the noble beasts like this will one day return…</p>
<p> </p>
<p><strong><em>Key Lessons: Hardware</em></strong></p>
<p> </p>
<ul><li>Buy core system-backbone components as late as possible in the process, and the graphics cards even later.</li>
<li>Be dutifully sceptical about the cost-benefit of Intel’s premium enthusiast-grade flagship processors.</li>
<li>Core count is slightly more important to me than clock speed, but both must be considered carefully.</li>
<li>I don’t need as much RAM as some of you seem to.</li>
<li>A <em>GTX Titan</em> model will likely have only token superiority over the soon to follow <em>Ti</em>.</li>
</ul><p> </p>
<p><strong>Cooling Loop</strong></p>
<p> </p>
<p><em>Ironbeast</em>’s crosslinked cooling loop design is probably the most technically complex of its experimental ideas. Its intention was to allow all the system components to have shared access to the full cooling capacity of the system, as in single loop, whilst retaining the reduced restriction advantages of a dual loop configuration. This would allow for the maximum usage to be gleaned from all the available radiators all of the time, no matter what the machine is doing – be it a CPU-heavy application, GPU-heavy application, or taxing to both. I am delighted to report that the concept does indeed appear to work effectively, though it does have its eccentricities. <em>(It might even suggest that a little of the knowledge garnered from university is actually still in here somewhere. Maybe it wasn’t a total waste of time after all…</em>).</p>
<p> </p>
<p>Likewise the smooth quad-GPU coolant split using the manifolds also works effectively, despite my concerns about adequate throughput running a four-way vertical separation and reintegration of the flow. However the use of the <em>EK</em> waterblocks – with their lovely appearance but conventional port arrangement - rather than the <em>Alphacool NexXxos GPX</em> – side-ported, like my concept – was a mistake. It forces an unnecessarily complex arrangement of right-angled turns, something which - engineering degree talking - I strongly wanted to avoid. This made all the effort in producing a smooth split completely pointless and rendered the manifolds a pretty but useless affectation. Similarly this choice prevented the Aqua Computer MPS flow sensors from fitting in between the cards <em>(or more accurately, the sensor's screw heads and protruding cabling)</em> prompting me to scrap the rig’s proposed extensive monitoring system. <em>(Sad about that, but in truth the plan only really existed to satisfy my intellectual curiosity about Ironbeast’s thermodynamic properties.)</em></p>
<p> </p>
<p>The extensive use of fittings to construct entire pipework sections, plus the centre tubes in the reservoirs, has allowed for much of the required pipework to be squeezed into a space which simply would not be possible otherwise – essential in producing <em>Ironbeast</em>’s signature symmetrical layout. It is however a murderously expensive method, only really rendered necessary by the space constraints of the 900D. I would make similar use of fittings again in future, but not to nearly the same extent. Bitspower in particular cannot be praised highly enough for their quality, and Monsoon’s 45-degree lightport fittings should also be mentioned as the thing that made those manifolds possible.</p>
<p> </p>
<p>Although the loop design has been successful, there are a number of areas where the concept possesses scope for revision and improvement. Putting the coolant returns at the tops of the reservoirs to produce the illuminated jets of bubbles works very nicely <em>(as you can see from the recent pictures and .gif)</em> though it is extremely temperamental about how those bubbles manifest at any particular moment. <em>(Sometimes there are lots, sometimes almost none, sometimes large, sometimes downright frothy, and despite running the loop for ages now I still can’t work out why this is).</em> Nonetheless this system does have a problem. Much of the CPU-side pipework and the top radiator is positioned above the water level of the reservoir; causing backflow under gravity when the pumps stop, reinduction of air into the last part of the loop pipework, and limits the tanks to being no more than about 60% full during operation. One-way valves added in later were effective, but they proved to be more trouble than they were worth. I really should have anticipated this issue when designing <em>Ironbeast</em> in the first place, and it is something which will have to be carefully considered in future.</p>
<p> </p>
<p>The adaptation of the <em>Ironbeast</em> design to use D-plugs was an excellent, frankly sanity saving decision; the loop as originally planned would have been far more awkward to assemble. Nonetheless the system does seem particularly prone to getting pockets of trapped air during and immediately after filling – something which is easy to vent in certain places and extremely frustrating in others. Any subsequent projects that I do should incorporate otherwise superfluous T-junctions to provide venting points at key places, and make greater use of quick-disconnect fittings to facilitate easier maintenance. Where possible the waterblocks should be placed at the lowest point of the relevant loop section, the connecting pipes running down to them rather than up, in order to facilitate air clearance. In general my main conclusion for upcoming projects would be that the crosslinked design is better reworked toward a multiple interlinked loop structure - and that the cooling system would likely benefit from the overall configuration becoming more complex in order to make individual subsections simpler.</p>
<p>  </p>
<p><strong><em>Key Lessons: Cooling</em></strong></p>
<p> </p>
<ul><li>Radiators are more effective than I feared they might be.</li>
<li>Broken-stack is effective if the spacing is right, but still less than preferred.</li>
<li>Dust filters are very necessary in here.</li>
<li>The split-return bubble generator idea works well, if eccentrically.</li>
<li>I must pay more attention to potential backflow issues with top-entry loop returns.</li>
<li>I’m not nearly as sensitive to fan noise as some people.</li>
<li>Don’t overestimate the pump power necessary for adequate throughput in parallel branches.</li>
<li>Don’t underestimate the pump power necessary for tight, complex pipework arrangements.</li>
<li>D-plugs and quick-disconnects are very useful.</li>
<li>Provide additional vent positions and strategic break points to aid loop maintenance.</li>
<li>Practice how to do pipe-bending before my next build.</li>
<li>A much more complex loop comprised of individually less complex sections may give better utility.</li>
<li>Purchasing <em>Ironbeast</em>’s fittings provided 99% of all UK retail revenue for the financial year 2015-16.</li>
</ul><p> </p>
</div>
<a class="spoiler-link H-spoiler-toggle" href="#"><strong>Reflections on Aesthetic Aspects: Thoughts on Ironbeast's appearance</strong> <span class="spoiler-help">(Click to show)</span></a><div class="spoiler-hidden">
<p style="text-align:center;"> </p>
<p style="text-align:center;"><span style="font-size:18px;"><span style="text-decoration:underline;">Aesthetic Aspects</span></span></p>
<p> </p>
<p>Looking back it would be fair to say that the aesthetic elements of the build have been by far the most unambiguously successful part of this project. When I was working on the earliest designs for what would eventually become <em>Ironbeast</em>, back in the spring of 2014 when all this was still a pipe-dream, the prevailing fashion was for dramatic high-contrast black & white designs illuminated with white LED lighting. The whole idea of adjustable RGB lighting that seems so popular now was still very uncommon in custom builds – RGB LED strips and controllers were available, but there certainly was not yet the smorgasbord of components which now integrate flexible lighting colours as a matter of course. Likewise cases tended to be very over-styled and massively attention seeking <em>(most of the retail offerings)</em> or rather stark and utilitarian <em>(like Caselabs)</em>. Hardly any seemed to have the potential for the really elegant look that I so wanted - the subtle contrast between matt black, gloss and the dark brushed aluminium, the silver highlights, the frameless glass and a full-size window panel covering the entire side. The Corsair 900D was the only one that I felt could get close to it, in the size that was needed, and that of course is how this all began…</p>
<p> </p>
<p>A lot has changed since then. Now we find that pretty much everything includes basic RGB functionality as if it were somehow obligatory – fans, motherboards, GPUs, fittings, coffee mugs, cakes…the lot. Furthermore the subtle dark brushed aluminium and frameless glass look that so appeals to me has – whilst I wasn’t looking – seemingly become rather de-rigueur of late, with a lot of the major manufacturers offering their newer cases in that style. The ridiculous but inescapable conclusion is that the <em>Ironbeast</em>-look has become…well, <em>fashionable</em> - and that I might just have done something...<strong><em>trendy</em></strong>??? Well there’s a first time for everything I suppose…</p>
<p> </p>
<p>All told I am very pleased by how the styling choices worked out, and hopefully you agree. The use of the gold fittings as an accent colour, to add visual interest when the rig isn’t lit, has proved very successful. That said, while I love the Monsoon fittings – they look superb – in retrospect they may have been the wrong choice. The soft tubing ones are excellent, the <em>Hardline</em> versions less so; the plastic end-cap design is very hit and miss. Perhaps it’s just down to my inexperience, but the Bitspower type <em>(which was the alternative)</em> looks like it would have been easier to use – even if it does demand 12mm rather than 13mm outer-diameter tube. Doubtless others will be better informed on that.</p>
<p> </p>
<p>On other aesthetic matters the dark brushed aluminium effect wrap was, on the whole, very successful. It gets tricky if you try to go round any edges, but on a flat face and cut to the profile with a scalpel it was very easy to use. It doesn’t truly give the same effect as real brushed aluminium when you have the benefit of a direct side-by side comparison, but otherwise it does make for a surprisingly respectable imitation. Similarly the use of tile trim to edge the case panels has also proved to work very well, though it is a little more untidy at the corners than my compulsive tendencies would like. When it comes to cutting and drilling acrylic, fellow novices should be advised that at 3mm thick the sheet seems very brittle and prone to stress fractures, and even with the protective backing it will pick up surface scratches very easily, especially when trying to work in rather ad-hoc conditions. For more advice, consult the many people here who actually know what they’re doing…</p>
<p> </p>
<p>As for the matter of spray painting, a word of caution. If you are using it on unfinished wood, or a composite like the Trespa, you must be prepared to do the donkey work if you want it to look right. Such materials will naturally absorb at different rates in different places; so even when using quality stuff on it, multiple coats of primer, paint and repeated sanding back will be required to make a good job of it. Protective clear topcoats and lacquers <em>will</em> change the perceived finish quite dramatically. Furthermore I would advise people that it is VERY difficult to get rattle-can painted stuff to look good when placed next to real metal surfaces. Again good quality paint and careful preparation will get you a very respectable metallic lustre, enough to go with matt or brushed metal, but it will not look as good as the polish on your fittings. There certainly are ways to do mirror-finish on plastic – albeit a little more complicated on certain surfaces like fan blades – using airbrushes, gloss back base coats, chrome paint/powder, lacquer and pigments. However you really need a proper workspace to do that kind of thing, something that lamentably I do not have.</p>
<p> </p>
<p>The custom lighting setup is one of the more interesting elements of <em>Ironbeast</em>’s aesthetics, and there are several things to comment on here. Firstly the use of the side-glow optical fibre and lighting strips instead of individual LEDs works brilliantly, highly recommended. It gives a great effect, saves a lot of hassle with wiring, it’s easy to order online in varied thicknesses and lets you have an RGB effect anywhere designed for a standard 3 or 5mm push-fit LED. <em>(Like the XSPC Raystorm or the Alphacool Cape Fuzion Cores that I have in here, and indeed almost all other watercooling components with Plexiglas).</em> It is definitely something I shall look to use again in any forthcoming new rig.</p>
<p> </p>
<p>With regard to the fan backlighting, well that was less than easy to achieve but worth it. The strips used here were a pain to solder the wires to, which were then difficult to place neatly, but the end result is really rather beautiful. The <em>Corsair SP-120</em>s, with their broad flat blades, definitely give a better effect than the front ones – although the brightness does have to be set just right, as does the rotation speed with the front <em>AF-140</em>s. It would have been nice to be able to obscure the LEDs themselves, so you just got the backlighting effect, but dimensionally that simply isn’t viable. All that being said, it is rather unlikely anyone will bother with this sort of thing again; not now that RGB fans are so commonplace <em>(aesthetically the Thermaltake Riing is my current favourite)</em>, even if pre-coded effects are not quite as versatile as my Neopixel strips. I shall have to try out some of the new RGB products before my next project, to see in what ways they might compliment some of the custom stuff I developed for <em>Ironbeast</em>.</p>
<p> </p>
<p>On that subject, there is the matter of standard RGB and addressable. As you know my system uses <em>Adafruit Neopixel</em> strips run by an <em>Arduino Mega</em> microcontroller. As such every single LED can be set individually at any given moment, unlike a normal RGB strip, which lets you develop your own mesmerising yet subtle brightness and colour shifting effects that I have barely dipped a toe into experimenting with. Being powered by USB and independently controlled the system can be lit even when the beast isn’t running, and without the constraints of propriety software. On the flip side it’s much more involved in its wiring than simply plugging something in, and you don’t get the luxury of an easy, convenient, GUI application to play with – you have to code <em>(even if it is in a horribly amateur way)</em>. That said, although this was the case when I started, the <em>ASUS ROG Rampage VI Extreme</em> – the board that everybody will end up using, to judge by the <em>IV</em> & <em>V</em> – recently drew my attention, in that it lists on its internal I/O specification an addressable RGB header. So with these kind of LEDs run straight from your motherboard, plug-and-play RGB components, and stylish frameless glass cases straight off the shelf, anyone who wanted to follow the <em>Ironbeast</em> look should find it far, far, easier than it was for me.</p>
<p> </p>
<p>And if even one little thing from this build inspires someone, then I truly couldn’t be more proud...</p>
<p> </p>
<p><strong><em>Key Lessons: Aesthetics</em></strong></p>
<p> </p>
<ul><li>Custom RGB lighting for PCs is now much easier and more commonplace than it was for me.</li>
<li>The addressable type of RGB strip is starting to become a much more viable proposition too.</li>
<li>Sideglow optical fibre works brilliantly; both for looks and to eliminate messy wiring to individual LEDs</li>
<li>For those doing conventional layouts a nice range of stylish glass-sided cases are now available off the shelf.</li>
<li>Don’t ask me about cutting acrylic, I’m useless at it.</li>
<li>Using compression/hardline fittings in a different accent colour can produce a very nice extra level of detail.</li>
<li>Hex mesh looks much better than circular or rounded-slot mesh</li>
<li>Brushed aluminium effect wrap is a very good alternative to the real-deal, spray paint or to powder coating.</li>
<li>Creating a proper mirror-finish on plastic components is a very involved proposition.</li>
</ul><p> </p>
</div>
<p> </p>
<p style="text-align:center;"><span style="font-size:18px;"><span style="text-decoration:underline;">Final Thoughts</span></span></p>
<p> </p>
<p>With this build at its end, I can finally start working on planning a new machine. There are many new ideas which have emerged from this project, a few of which you have heard outlined here. Contrary to what I expected at the outset, there <em><strong>will</strong></em> be another build – the imperfections must be corrected! The earliest concept work for an <em>Ironbeast Mark II</em> began way back, since before the <em>Phase 3</em> parts were sent for manufacture and the design of this thing became fixed. However if there is one lesson I have learned from this, it’s that these things take a long time to do right. Therefore it’s probably going to be quite a while before any ‘<em>Ironbeast Evolution</em>’ build log makes its debut here. As I mentioned recently, I have been fleshing out a couple of options for an interim build (or builds) between now and then, mostly as experimental test beds for various elements of <em>Mk.2</em>.</p>
<p> </p>
<p>Given that <em>Ironbeast’</em>s hardware is already nearly three years old, and so more than half way through its expected useful lifetime <em>(at least as far as front line service is concerned)</em>, one proposition is to make another big power-rig to ultimately take over the heavy lifting. This would use the surplus 900D that <em>Ironbeast</em> didn’t, which has been sitting here looking lonely and unloved for far too long <em>(something I can’t stand)</em>. It would be a relatively conventional standard-layout but unconventional-loop build based broadly on my original <em>proto-Ironbeast</em> plan, and consequently is being worked on under the name <em>‘Precursor’</em>. However, as loathe has I am to leave my 900D empty in a corner for even longer, <em>Ironbeast</em> is still a very capable machine at present and I’m not wholly convinced that the latest batch of tech releases are going to be worthy of a successor PC. The alternative is a little micro-ATX or ITX project as an auxiliary and backup machine, either here alongside <em>Ironbeast</em> or for my bedroom. This would probably have a horizontal motherboard arrangement, be used to test the proposed custom case construction and aesthetic ideas for <em>Mk.2</em> on a smaller scale, and is perhaps a better proposition for what to do next. It is being currently being developed under the name <em>‘Timberwolf’</em> - which gives you a fairly big hint as to the direction I’m thinking of going with it.</p>
<p> </p>
<p>In either case there will eventually be a new build log for you to follow – which will probably start at a similar sort of point in the design process to <em>Ironbeast</em>, when there is a model and some nice new renders for you to look at. I trust you’ll take a glance at that one too.  </p>
<p> </p>
<p> </p>
<p style="text-align:center;"><em>This has been a very long road, and not a smooth one, but as a closing remark it must be reiterated one last time just how much I have enjoyed doing this build log for you all - and more than that, how much I have enjoyed that <strong>you</strong> have enjoyed it. Being a complete novice led me to be very hesitant in choosing to put this project out there, indeed in joining OCN in the first place. It is only your continuing enthusiasm for this build that got us this far, fuelling a compulsive need not to let you down. It may not be quite the machine that was promised, but nonetheless I hope I haven’t.</em></p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"><em>To everyone here who has read, subscribed, commented and generally been so much nicer than expected:</em></p>
<p style="text-align:center;"> </p>
<p style="text-align:center;"><em>Thank you all</em></p>
 

·
Chief Over-Engineer
Joined
·
608 Posts
Discussion Starter #1,307
<div class="quote-container" data-huddler-embed="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26297828" data-huddler-embed-placeholder="false">Quote:
<div class="quote-block">Originally Posted by <strong>nycgtr</strong> <a href="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26297828"><img alt="View Post" src="/img/forum/go_quote.gif" style=""></a><br><br>
Just drain it and replace the board and cpu lol. Be a whole lot easier.</div>
</div>
<p> </p>
<p>You think so? That was always the single biggest weakness with the <em>Ironbeast</em> design – it’s so highly tailored to this specific hardware. I knew from the start that would be the compromise. </p>
 

·
WaterCooler & Overclocker
Joined
·
2,581 Posts
I just finished looking at all the final pictures and reading through your "Project Evaluation" post. Awesome pictures of a magnificent machine!<br><br>
I know you posted the final pictures several weeks ago, but I have left it sitting in my subscription page until now, because in a way I just did not want to see this build come to an end. I've been following along since the beginning and have always looked forward to seeing your progress/updates.<br><br>
Early on when you were showing renders I thought to myself "there's no way this guy is going to be able to pull this off". It took a long time, but it was well worth it. You certainly pulled it off with style!<br><br>
You did so many unique things on this build that I've never seen before, like the LEDs in the fans, the awesome looking reservoirs to the rollers on the bottom. My favorite part of the build is the water manifolds to the GPUs... just spectacular. I'm sure if there is an award for "Most fittings used in one build" you would win that for sure <img alt="biggrin.gif" class="bbcode_smiley" src="http://files.overclock.net/images/smilies/biggrin.gif"><br><br>
Thank you so much for sharing your work here. I have thoroughly enjoyed following along and will look forward to any new project you start. I know I've repped you several times along the way, but I'm giving you another rep today for actually finishing this fabulous build.<br><br>
It is one of my favorite builds of all time <img alt="thumb.gif" class="bbcode_smiley" src="http://files.overclock.net/images/smilies/thumb.gif">
 

·
Chief Over-Engineer
Joined
·
608 Posts
Discussion Starter #1,310
<div class="quote-container" data-huddler-embed="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26318092" data-huddler-embed-placeholder="false">Quote:
<div class="quote-block">Originally Posted by <strong>Barefooter</strong> <a href="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26318092"><img alt="View Post" src="/img/forum/go_quote.gif" style=""></a><br><br>
I just finished looking at all the final pictures and reading through your "Project Evaluation" post. Awesome pictures of a magnificent machine!<br><br>
I know you posted the final pictures several weeks ago, but I have left it sitting in my subscription page until now, because in a way I just did not want to see this build come to an end. I've been following along since the beginning and have always looked forward to seeing your progress/updates.<br><br>
Early on when you were showing renders I thought to myself "there's no way this guy is going to be able to pull this off". It took a long time, but it was well worth it. You certainly pulled it off with style!<br><br>
You did so many unique things on this build that I've never seen before, like the LEDs in the fans, the awesome looking reservoirs to the rollers on the bottom. My favorite part of the build is the water manifolds to the GPUs... just spectacular. I'm sure if there is an award for "Most fittings used in one build" you would win that for sure <img alt="biggrin.gif" src="http://files.overclock.net/images/smilies/biggrin.gif"><br><br>
Thank you so much for sharing your work here. I have thoroughly enjoyed following along and will look forward to any new project you start. I know I've repped you several times along the way, but I'm giving you another rep today for actually finishing this fabulous build.<br><br>
It is one of my favorite builds of all time <img alt="thumb.gif" src="http://files.overclock.net/images/smilies/thumb.gif"></div>
</div>
<p> </p>
<p>‘All good things…’ as they say. Though in truth that has always seemed an appallingly defeatist maxim to me.</p>
<p> </p>
<p>Sad that we had to bring it to a close, but it was long past time – though doubtless you lot could probably have talked me into carrying on forever if you wanted to. It is genuinely heart-warming to know how much you’ve enjoyed reading the build log, and how much you like the final result. Very kind of you to give the +Rep too. There’s something especially gratifying about having someone who was here at the beginning still here at the end, telling me it was all worthwhile.</p>
<p> </p>
<p><em>(Seriously though, how on earth have you put up with me for so long! I’ve completely lost track of how many different builds TCO has completed in the time it’s taken me to do this. Four? Five?)  </em></p>
<p> </p>
<p>All this being the case, and since you didn’t want to see this end, it should probably be mentioned that <em>Ironbeast</em> has its routine drain down and coolant change pencilled in for some point in the next week. I may have thought up an alternative solution to at least partially fix the backflow problem and the temptation to start messing around with it again seems to be getting the better of me. Yes, I know – the build is over. Done. Finished. No more working on it, start on something new. But if successful this particular modification should make the whole machine run so much more smoothly, and one little curtain call wouldn’t do any harm…right?</p>
<p> </p>
<div class="quote-container" data-huddler-embed="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26319600" data-huddler-embed-placeholder="false">Quote:
<div class="quote-block">Originally Posted by <strong>m1ndb3nd3r</strong> <a href="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1300#post_26319600"><img alt="View Post" src="/img/forum/go_quote.gif" style=""></a><br><br>
Love this build... Just finished reading it all,</div>
</div>
<p> </p>
<p>Much appreciated, glad you like it. But surely you didn’t read <em>all</em> of it? My ponderous rambling walls of text?</p>
 

·
Registered
Joined
·
293 Posts
Soooo.... decided on whats next <img alt="biggrin.gif" class="bbcode_smiley" src="http://files.overclock.net/images/smilies/biggrin.gif">
 

·
Chief Over-Engineer
Joined
·
608 Posts
Discussion Starter #1,313
<div class="quote-container" data-huddler-embed="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1310#post_26406035" data-huddler-embed-placeholder="false">Quote:
<div class="quote-block">Originally Posted by <strong>Gleniu</strong> <a href="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1310#post_26406035"><img alt="View Post" src="/img/forum/go_quote.gif" style=""></a><br><br>
Fantastic build, looks amazing. A great summary and thoughts - thank you for sharing. <img alt="thumb.gif" src="http://files.overclock.net/images/smilies/thumb.gif"></div>
</div>
<p> </p>
<p>You’re far too kind. I’m glad <em>Ironbeast</em> was worth sharing - despite taking years for me to drag it into reality.</p>
<p> </p>
<div class="quote-container" data-huddler-embed="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1310#post_26403989" data-huddler-embed-placeholder="false">Quote:
<div class="quote-block">Originally Posted by <strong>Angry-Hermit</strong> <a href="/t/1544054/build-log-ironbeast-a-fully-water-cooled-x99-quad-gpu-900d-build/1310#post_26403989"><img alt="View Post" src="/img/forum/go_quote.gif" style=""></a><br><br>
Soooo.... decided on whats next <img alt="biggrin.gif" src="http://files.overclock.net/images/smilies/biggrin.gif"></div>
</div>
<p> </p>
<p>I have indeed...</p>
 
1301 - 1313 of 1313 Posts
Top