Synthetic Benchmarks
3DMark Firestrike
Straight away, the Threadripper is off to a very strong start. The higher clock speeds of the i9 do seem to be giving it a lead, but without a doubt, the 1950X is competitive with Intel’s top dog.
Stock
Overclocked
Overclocking did bring some small boost to the Threadripper, but Intel holds the lead here, again most likely due to their higher clock speeds. Threadripper is seriously hot, and unless we can get a bigger cooler, we can’t clock it higher.
Unigine Superposition 1080p Extreme
No shot at the top spot for the 1950X here, however, the score is competitive, and there’s not much deviation between the top and bottom of the charts here. The 1900X didn’t score that much higher either.
Stock
Overclocked
Overclocking did tighten up the performance nicely, closing the gap between all the CPUs, and even scoring a tiny bit higher than the 1900X. Of course, the scores are so close, you likely wouldn’t notice any difference here, as this is a very GPU bound task. Either way, it shows that PCIe throughput is unhindered.
PCMark 10 Express
Now, this is something I did not expect. Multitasking is amazing on the Threadripper, but the higher clock speeds of the Intel chips is still king here. Of course, the 1950X is no slouch at all, but this is still a very interesting result.
Stock
Overclocked
Overclocking didn’t bring many benefits again, and this tells me one big thing, that PCMark 10 just isn’t using all those cores to their fullest. Close speed on a per core basis rules the roost in this test.
WPrime 32M and 1024M
The 1950X isn’t built as a day-to-day gaming CPU, it just isn’t. However, when it comes to heavy duty workloads such as rendering and calculations, it’s in a league of its own. It decimated the i9’s time of 67.946 by 10 seconds! Run this test on your own system, see how close you can get to that time… simply amazing performance here.
Stock
Overclocked
A reasonable overclock helps a fair bit here too, cutting that time down to a record shattering 48.986 seconds, a time that we don’t think will be beaten in a hurry.
Cinebench R15
Again, heavy duty rendering and computational workloads are what Threadripper is all about. Immediately, we can see it walks all over the i9-7900X’s score. Don’t get me wrong, the i9 is a monster at these tasks too, it’s no slouch. However, cores are king when it comes to rendering and with a score of almost 3000, the 1950X is almost twice as fast as the 1800X!
Stock
Overclocked
That performance is taken to the next level by overclocking too. The i9 didn’t gain much, but the 1950X boosted by over 400 points in this test. If you’re running Blender, Premier, or any rendering application, this CPU is going to be a literally time saver.
Handbrake MP4 to MKV Conversion 4K
Interestingly, the i9-7900X kept the lead here, but the 1950X isn’t exactly lagging behind much. Both CPUs put up a strong fight in this test. Shame we couldn’t keep the 1950X cooler to get those clocks that bit higher, as it may have nipped first place here.
Stock
Overclocked
Overclocking actually saw the 1950X perform worse in this test. I attribute this to overheating half way through the test. Rest assured, we’ll be investing in a big cooler and retesting both the i9 and the TR in the coming weeks to see which one really is the king at rendering.
The gaming performance is shocking. Something isn’t right when Ryzen chips are doing better. Probably some microcode improvements to be made me thinks.
Anyway, this is what Ryzen should’ve been!
Did the review using ThreadRipper’s Gaming Mode for the gaming tests? See AnandTech…
IMO you’re not 1080p gaming with a TR if you’re even slightly sane, so those results can be ignored.
Why should 1080p be ignored? Lower resolutions mean higher CPU work.
Oh come on, even kindergarten kids know that ThreadRipper supports ECC RAM. Please fix that in the article.
https://www.amd.com/en/products/ryzen-threadripper
“Up to 16 cores and 32 threads for lightning-fast creative workloads.
An unprecedented 64 PCIe® Gen3 lanes to meet large GPU and NVMe needs.
Up to 40MB of combined cache for rapid access to large data sets.
Quad channel DDR4 with support for ECC for reliable throughput.”
This processor is really huge and so is its raw performance. It is quite shame that the gaming performance is not at all satisfactory.
“For some gamers, not having a couple of FPS extra isn’t a big deal, especially if the CPU can still render a video stream in the background without making other compromises. It’s tough to say how many, but there’s a market for a CPU like this, it’s likely just not your typical PC gamer.”
Seems a convoluted pro intel argument.
How about:
If the knowledge worker Cpu CANT still render a video stream/compile etc.in the background without making other compromises, then It’s tough to say how many, but there’s a market for a CPU like this, it’s likely just not your typical user in control of their life.”
The question for the target market is?, can it excel at work tasks, yet perform a dual task, OR, is it much worse than a dedicated, separate, single app, futureless gaming pc. To say “you get no better gaming for the extra money is irrelevant.”
If both needs exist, but priorities are reversed, then clearly the ryzen suits, esp the frugal 6 cores for some reason, and a little more patience is needed for the infrequent core heavy tasks. (Sigh, it seems only yesterday 8C WAS for core heavy tasks.)
Are people also really as silly as pundits assume, & they think they are dealing with a static target – that their PCs wont get loaded up with more and more distractions over the years of ownership?
Pundits also fail in their duty to newbies by not forewarning of the inevitable lane shortage they face when they upgrade meaningfully. Only a $1000usd Intel or TR give any real latitude/peace of mind on expanding HB resources.
They should also be warned (long story) that intels ~switched/filtered lanes are not the equal of amdS honest lanes.
Some loss of latency from doubling ryzens zeppelin dies on a a TR MCM is inevitable. Whats remarkable is how little a latency penalty is paid for the benefits of extra lanes. They still overclock to the ryzen sweet spot of 3200 ram & 4k OC for the cpu.
I am puzzled by memory tests here. Others have shown? a ~linear benefit from the 4 channel TR vs the 2 channel ryzen – a major bonus if true?