We’ve been exploring this really silly honkin’ great big dual Xeon workstation build recently, but I don’t think we’ve quite hit the limits of its potential with the 1080Ti. Also, I want a better handle on the difference, if any, that all those extra cores on the second CPU make in gaming.
Now, obviously gaming is very much not the use case for a system like this, but here’s the secret – they can’t stop us doing it. No-one can stop us. There are no barriers any more. Embrace the chaos. I’m told it’s a ladder.
Welcome to Esk Computers, I’m Scott and today on the desk of Esk, let’s see if we can’t answer those questions.
A couple of weeks back we tested out a single Xeon E5 2680 V4 build to see what it can deliver when paired with a 7800XT graphics card, which is the fastest GPU I have access to. Seeing as we’ve just built a workstation cum server with two E5 2680 V4’s, it seems like a logical idea to see if that extra processor unlocks any extra performance in gaming workloads, or if they wind up tripping over themselves when trying to feed the 1080Ti.
To cut a long but not visually interesting story short, I’ve removed one of the CPUs and rerun the 1080Ti tests seen in the build video last week to see the difference that makes, if any. No, not that one, it won’t boot with that removed. Yes, that one, that’s more like it.
Then, I’ve reinstalled that CPU, wiped the drivers, and popped in the 7800XT. We can then compare that to the results with my 9800X3D to see what performance the older Xeons leave on the table. It would be cleaner as two separate videos, I suppose, but that would be a bit content-light. So, I will trust that this data deluge isn’t too confusing, and crack on.
I’m only really including the Cinebench 2024 Multicore test to ensure there’s at least one test where the second CPU makes a difference. As you see, the double Xeon setup’s not exactly twice as fast as the single CPU, at 706 vs 1157, but still a big ol’ number and within touching distance of the 9800X3D, if you define 11% as touching distance, although at 16 threads vs 56, maybe it’s not as impressive as it looks.
Firing up 3DMark now, and in Fire Strike we’re seeing a 10% difference, although it’s in the favour of the single Xeon setup in the 1080Ti tests, at 21486 vs 19351 points. The 7800XT result is down 53% on the 9800X3D results and at 20231 points, not actually faster than the single CPU and 1080Ti result.
There’s an even more stark difference in Time Spy, where the single Xeon and 1080Ti is 24% faster than the dual CPU machine, at 9897 vs 7541 points. A wild margin, but retesting gave the same results. Again, the 7800XT score of 7753 in not only 59% down on the 9800X3D result, it’s also about 22% lower than the single CPU and 1080Ti score.
The last of the synthetics now with Superposition at 1080p High settings, where the single CPU and 1080Ti is technically faster at 14390 vs 14017 points, but that’s margin of error levels of difference. At least the 7800XT result is for once faster than the 1080Ti and single CPU combo, by a whopping 3% at 14796. That’s 37% down on the 9800X3D result.
Shadow of the Tomb Raider’s benchmark at 1080p High shows a barely perceptible win for the dual CPU and 1080Ti setup at 132 vs 130 average frames per second. Because apparently this build is haunted, the 7800XT and dual CPU scores 9% lower at 119 fps. That’s a 55% drop from my 9800X3D result.
The haunting continues with Counter-Strike 2, where the results on the dual CPU system with the 1080Ti are just weird. There’s no appreciable difference with the Very High settings results at 123 vs 125 average fps, although dropping to the Competitive settings, Low, but with all dynamic shadows, we get 149 fps with one CPU and 109 fps with two. Both setups seem like they should score higher, and I still can’t find a reason for the dual CPU dropping so much performance when dropping settings. Just bizarre, but re-testing confirms it.
The dual Xeon and 7800XT results show barely any improvement, with the Very High settings run giving 152 fps, but at least the Competitive settings obeys logic and reason by increasing, slightly, to 161 fps. That’s a 59% drop from the 9800X3D results at Very High, and a monstrous 80% drop at Competitive settings. Just plain weird. I’m going back to Global Offensive.
Cyberpunk 2077 returns to comparative normality, with the inbuilt benchmark at 1080p giving practically identical results on the 1080Ti regardless of CPU quantity at 75 and 74 average fps. There’s a roughly 9% increase in going to the 7800XT, which is a 53% drop from the X3D results at 82 vs 175 fps.
Taking control and running around the streets of Kubuki, and harshing the mellow of the pigs gives an edge, arguably, to the single Xeon and 1080Ti setup over the dual CPU at 53 average fps vs 49 fps, although again that’s also mostly a margin of error difference. The dual Xeons with the 7800XT show no improvement, remaining at 49 fps, that’s a 60% drop off from the 123fps the 9800X3D system gave me.
Black Myth Wukong’s benchmark at 1080p High, with no Framegen and no raytracing also shows no difference with the single and dual Xeon and 1080Ti setups, both getting 38 fps. For once, there’s a sizeable 92% increase with the 7800XT to 73 fps, which isn’t that far off the 83 fps from the 9800X3D, 12% in fact.
More Unreal Engine 5 action with Marvel Rivals, where the benchmark at 1080p Medium, Native Resolution, no raytracing again gives a very marginal victory for the single CPU setup in the 1080Ti tests, at 60 fps vs 57 fps. There’s about a 40% increase with the 7800XT at 84 fps, which is about 28% lower than the 9800X3D’s 116 fps.
The Assassin’s Creed Mirage benchmark at 1080p High Settings is a wash in the 1080Ti testing, both dual and single CPUs getting 90 fps. The 7800XT gives us 130 fps, 44% better, although that’s 29% off the X3D’s 182 fps result.
Horizon Zero Dawn’s benchmark at 1080p Favour Quality settings gives a 5% win for the single Xeon setup with the 1080Ti, at 110 vs 104 fps. There’s a 24% jump to 136 fps by adding the 7800XT, again that’s a good ways away from the 218 fps the X3D can get you, 38% actually.
Baldur’s Gate 3 at 1080p Ultra settings next, and with the usual run around Lower City there’s a weirdly high 29% difference of 55 fps vs 39 fps, again in favour of the single CPU setup. It drops even further to 36% difference at 35 fps when adding the 7800XT into the mix. I know this test is a little less repeatable than the standard benchmarks, but I did run them again to confirm with more or less the same results. It just doesn’t like dual Xeons. The X3D gave 167 fps, a near 80% difference.
I’d have expected a win for the dual CPU setup in Total War: Warhammer III’s Mirrors of Madness benchmark at 1080P Ultra Settings, but au contraire, it’s beaten out by 33% with the single CPU result of 48 fps vs 32fps. The 7800XT gives a functionally identical low result of 31 fps, which is a sizable 73% drop from the 113 fps the X3D can give.
Forza Horizon 5’s benchmark at 1080p High settings again sees the single CPU setup edge out a 9% win at 102 fps vs 93 fps on the 1080Ti testing. There’s a 25% gain in adding the 7800XT, up to 127 fps, which is 51% down on the X3D’s 261 fps.
Red Dead Redemption 2 at 1080p favour quality settings is a wash, really, at 66 vs 65 fps for the 1080Ti testing. There’s a 61% boost in adding the 7800XT, to 107 fps, which is 54% down on the 235 fps that the 9800X3D will give you.
And we’re rounding off with a rare win for the dual Xeon setup in the 1080Ti testing of Civilization VI’s Gathering Storm AI Benchmark at 1080p Highest settings, with an average turn time of 46 vs 49 seconds. A 6% win. Hooray. Tempered only by this slowing down when adding the 7800XT to 53 seconds, about 13% longer. The 9800X3D took 24 seconds in this test.
So, welcome again to another episode of Scott proves the bleedin’ obvious, that a system based on decade old dual Xeon technology is not at the cutting edge of gaming performance. I still think there’s a bit of life left in the single Xeon CPU configuration – but don’t build something like this if it’s purely for gaming.
Arguably for max performance, I should be testing the 7800XT with only one of the Xeons, but one does not buy a dual CPU board to use only one CPU, does one. Besides, we’ve already basically done this previously.
It’s not an exact comparison given the different motherboard and whatnot, but we did test a single Xeon 2680V4 with the 7800XT a few weeks back. Check that video out for the full results. I’ll spare you chapter and verse on it here because it’s really the same story we saw here with the 1080Ti, the single CPU machine just performs better in gaming workloads than the dual setup. The best it can hope for is parity.
I expect the overhead in shuffling data between the CPUs is causing the inefficiencies. I think there’s a route to reclaiming the difference in performance without yanking out the processor by using a tool like Process Lasso to restrict games into running on one CPU, but seeing as this is not going to actually be a full time gaming machine, or indeed running Windows, I’m not going to worry about it too much.
When this is properly configured as a server I think it would be useful to have it set up in part as my own personal cloud gaming service, but that’s a ways down the line. When I’ve got some flavour of Linux on this we’ll need to investigate if it’s got the same problems with thread scheduling and if so how to fix them. But that’s a problem for a future video.
In the meantime, if you have any questions or want further details please leave a comment down below, and if you enjoyed this videotronic missive then consider subscribing. Until next time, take care of yourself, and each other.