Better Budget X99 Build!

Welcome to Esk Computers, I’m Scott and today on the desk of Esk, we have another budget v4 Xeon build, this time paired with a GTX 1660, on paper only a small step up from the RX580 we looked at recently, but in real-life, it’s punching well above its weight.

I call this Cursed Obelisk, because it has a dark history that was ultimately solved by swapping out a malfunctioning power supply and graphics card. As it stands now, it’s using one of these relatively cheap X99 (or so they claim) motherboards, in this instance a Machinist X99 paired with 2 8GB sticks of DDR4 2133MHz and the Intel Xeon E5 2680 V4, a 14-Core, 28-Thread part that clocks at 2.5 GHz with a turbo up to 3.3 GHz. This is a smidgen faster than the 12 core E5 2650 we just looked at, let’s see if that translates to better performance for not much extra money.

The single thread score in Cinebench 2024 is 45, which is 10% faster than the slower clocked 2650, although still less than half the speed of modern chips. Obviously, the two extra cores gooses the multi-core score to 708, which is within touching distance of parts like the Ryzen 7 5700X – not modern flagship performance of course, but still quite competitive for the right sort of application.

Comparison charts such as Tom’s Hardware’s Legacy list puts the GTX 1660 – we have a Zotac example here – only a little above the RX580 we looked at last time. Now, that was the cut down 2048SP version, so let’s see if a team green card that can be had for much the same price on the second hand market can outperform it.

Whistlestop tour of the other component first, this has a 3-fan LIONSONIC cooler with 6 Heat Pipes to deal with the theoretical 120W power draw, there’s a 512GB Partiot NVMe SSD that’s very fast, and a 2TB hard drive that is not. The PSU is a EVGA 600w unit, and the case is an iONZ KZ-35V Zenith , but it’s easy to build in and decent value, coming with three ARGB PWM fans which I’d repurposed for my own PC and swapped out for some common or garden unlit fans, seeing as there’s no ARGB headers on the motherboard and this is too simple a case to hide the mess of cables RGB fans and controllers bring.

Again, this motherboard has a fake lying USB3 header that’s wired to a USB2 controller, so I’ve added in this PCI card to give five ports at the back and a properly bespeeded header for the front case ports. This GPU is a bit thinner, so I’ve not needed to do the riser cable trick and the GPU is still keeping cool so doesn’t seem to be too affected by any reduction in airflow. I’ve let the NVIDIA app determine the overclock, so perhaps that’s leaving a little performance on the table but it’s the most I’d expect any non-enthusiast potential buyer to want to do.

In the not particularly interesting synthetic benchmarks, Time Spy gets 5878 points, Fire Strike 12512 points, and Superposition 7695 at 1080p High settings, increases of 35%, 12% and 44% over the 580 2048SP. Big numbers but let’s if that continues into the proper vidja games. 

Shadow of the Tomb Raider gets a very playable 88 fps at 1080p High Settings, a 35% improvement. Counter Strike 2 returns 95 fps at Very High settings, about 30% faster than the 580 SP. With numbers like these even though I haven’t tested them, I’m confident it’s going to run the likes of DOTA, League of Legends, Minecraft, Roblox, and Fortnite in performance mode very well indeed.

The harder to run stuff is where we will really cut some nuts. I have absolutely no idea why Cyberpunk 2077 would let me run turn on ray-tracing, apart maybe from as a marketing tool to get me to buy a card that’s capable of accelerating that workload properly. It gets a whopping 11 frames on RT Ultra settings, but without that newfangled RT nonsense it benches at 53 fps at 1080p Ultra settings, so in my opinion playable framerates with all the eye-candy enabled, or if you’d prefer, 65 fps and 1080p High settings. A great showing for the 1660, and on a like for like comparison with the 580 at 1080p Ultra settings it is 55% faster.

Unreal Engine 5 brings everything to its knees, and the 580 struggled greatly with it. The 1660 is numerically quite a bit better, but still needs some upscaling trickery to get into playable territory. Black Myth Wukong gets 25 fps at 1080p High settings – 38% better than the 580, but still not enough for this type of game. However, 74 fps at 1080p Low, and 65% TSR is much more like it, and all without framegen.

Marvel Rivals is also onboard the strugglebus, getting 35 fps for 1080p Medium settings without any upscaling, 66% better than a 580 but again, just not enough. To get to the threshold of playability, I’d suggest 1080p Low with Quality XeSS, which gets 63 fps without any latency adding framegen, albeit at quite a visual hit.

Taking a quick jaunt to Verdansk, Call of Duty Warzone hits 77 fps at 1080p Extreme settings. As I’m completely buttcheeks at this game I’m happy enough for it to just look pretty while I get murderised, but if you are a sweatier gamer than I then of course high framerates at lower settings are achievable. It’s about 70% faster than the 580 SP at these admittedly wonky settings

Horizon Zero Dawn is my representative for the poorly optimised PlayStation port contingent, where at 1080p High, Prefer Quality settings the benchmark demo gives 73 FPS average, which is about 23% faster than the 580 SP.

Forza Horizon 5 is a well optimised game, so as expected this runs well. The benchmark runs gives us an 81 fps average at 1080p High settings. That’s a 39% improvement over the 580SP’s 58 fps average, although as with most of games these we are getting very playable experiences out of both cards.

Assassin’s Creed Mirage at 1080 high settings is also running very well, the benchmark run netting a 63 fps average, another very solid performance.

Red Dead Redemption II’s benchmark run at the low end of its balanced settings at 1080p gives us an average frame rate of 59.5 FPS, a max of 80 and min of 49, which is about 11% faster than the 580SP, but I think that’s getting down in to the barely noticeable difference territory.

Now, I’m not the boss of you, and you are allowed to have your own opinions about what’s acceptable in 2025 as a gaming machine. This set of components isn’t going to be pushing out acceptable 4K performance in Unreal Engine 5 games, or anything with raytracing required. However it will cost something like a tenth of the price of the cutting edge hardware that needs and can give great 1080p results in the vast majority of PC games ever made.

In terms of directly comparing the 1660 and the 580 SP. well, as with all components it’s going to depend on what you’re able to pick them up for. If they were the same price, then the 1660 performs better across the board and draws less power doing it, but on the other hand I think it’s fair to say they are in the same performance class. By which I mean that I don’t think there’s any game that you’ll get a playable experience on the 1660 that you won’t be able to play acceptably on the 580 SP, albeit perhaps with slightly lower settings. So if you are cross shopping these cards on a tight budget and there’s a decent price difference between them, you’ll have a lot of success with the 580 SP, or even better, the full fat 580 version, or indeed the older RX 480 which is essentially the same card.

For me, while ideally I’d be looking to move on to AM4 and 20-series cards for budget systems, they will generally cost more than this can be put together for, exceptional used parts deals aside which I understand really varies in availability with where you are in the world. You’ll have to do your own comparison shopping, but unless you’re after playing the absolute latest titles I would still put this combination on the consideration list.