Menu
Intel’s high-end desktop dominance has been under siege ever since the name Threadripper was first uttered. The blue team’s, topping out, held a front against the first wave of AMD’s high-end Ryzens, but PC builders groused nonetheless—and rightfully so. Where did the solder that had joined heat spreaders to dies of past high-end Intel CPUs go, especially on a $2000 chip? Why did X299 motherboards have problems keeping their VRMs cool enough under extreme loads with the platform’s highest-core-count CPUs? Why did quad-core parts exist for the X299 platform at all?Next to the unrestricted, segmentation-free approach of Threadripper CPUs and the X399 motherboards, the X299 platform and the breadth of the CPUs that could light it up looked by turns stingy and scattered.
I9 (of Season 4) is a coordinate grid area of the Battle Royale Map that have been revamped with a map update. It is formerly located at the south west central part of Moisty Mire and was converted into a large production set with green screens, props and stages on Season 4. Points edit edit source Moisty Mire (South) Production Stage.
The Core i7-7800X and i7-7820X offered only 28 PCIe lanes from the CPU, compared to the 44 from the Core i9-7900X and better CPUs in the lineup. The entry-level Core i7-7800X didn’t even benefit from Turbo Boost Max 3.0, one of the headlining innovations of the Core X lineup. Worse, the 16 CPU-powered PCIe lanes from the Kaby Lake-powered Core i5-7640X and Core i7-7740X required motherboard makers to employ complex lane-switching schemes even on high-end mobos that seemed unlikely to ever play host to their four cores.Intel may have had good intentions in providing builders a wide range of choices and an upgrade path in putting together high-end systems, but the initial headaches of X299 suggested that strategy had stretched the platform a bit too far.AMD didn’t stand still with its high-end desktop CPUs in the intervening time, either. Didn’t just challenge the i9-7980XE in some workloads—it actually beat Intel’s highest-end desktop chip in some tasks for less money (though not in every test). Glancing though that blow may have been, the fact that AMD was even able to lay a finger on Intel’s high-end desktop performance crown was an indignity unimaginable just a couple of years ago. Our testing methodsAs always, we did our best to deliver clean benchmarking numbers. We ran each benchmark at least three times and took the median of those results.
Memory subsystem performanceThe AIDA64 utility includes some basic tests of memory bandwidth and latency that will let us peer into the differences in behavior among the memory subsystems of the processors on the bench today, if there are any.Some quick synthetic math testsAIDA64 also includes some useful micro-benchmarks that we can use to flush out broad differences among CPUs on our bench. The PhotoWorxx test uses AVX2 instructions on all of these chips. The CPU Hash integer benchmark uses AVX and Ryzen CPUs’ Intel SHA Extensions support, while the single-precision FPU Julia and double-precision Mandel tests use AVX2 with FMA. JavascriptThe usefulness of Javascript microbenchmarks for comparing browser performance may be on the wane, but these tests still allow us to tease out some single-threaded performance differences among CPUs.
As part of our transition to using the Mechanical TuRk to benchmark our chips, we’ve had to switch to Google’s Chrome browser so that we can automate these tests. Chrome does perform differently on these benchmarks than Microsoft Edge, our previous browser of choice, so it’s vitally important not to cross-compare these results with older TR reviews.WebXPRT 3is meant to simulate some realistic workloads one might encounter in web browsing. It’s here primarily as a counterweight to the more synthetic microbenchmarking tools above.WebXPRT isn’t entirely single-threaded—it uses to perform asynchronous execution of Javascript in some of its tests.
Compiling code with GCCOur resident code monkey, Bruno Ferreira, helped us put together this code-compiling test. Qtbench records the time needed to compile the Qt SDK using the GCC compiler.
The number of jobs dispatched by the Qtbench script is configurable, and we set the number of threads to match the hardware thread count for each CPU.File compression with 7-ZipThe free and open-source has a built-in benchmark that occupies every core and thread of the host system.Disk encryption with Veracrypt. CinebenchThe evergreen Cinebench benchmark is powered by Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. The test runs with a single thread and then with as many threads as possible.Blender. The app can take advantage of AVX2 instructions on compatible CPUs.
We chose the “bmw27” test file to put our CPUs through their paces.CoronaCorona, as its developers put it, is a “high-performance (un)biased photorealistic renderer, available for Autodesk 3ds Max and as a standalone CLI application, and in development for Maxon Cinema 4D.”The company has made a standalone benchmark with its rendering engine inside, so it’s a no-brainer to give it a spin on these CPUs.IndigoIndigo Bench is a standalone application, which creates photo-realistic images using what its developers call “unbiased rendering technologies.”V-Ray. HandbrakeHandbrake is a popular video-transcoding app. To see how it performs on these chips, we converted a roughly two-minute 4K source file from an iPhone 6S into a 1920×1080, 30 FPS MKV using the HEVC algorithm implemented in the x265 open-source encoder. We otherwise left the preset at its default settings.SPECwpc WPCcfdComputational fluid dynamics is an interesting and CPU-intensive benchmark. For years and years, we’ve used the Euler3D benchmark from Oklahoma State University’s CASElab, but that benchmark has become more and more difficult to continue justifying in today’s newly-competitive CPU landscape thanks to its compilation with Intel tools (and the resulting baked-in vendor advantage).We set out to find a more vendor-neutral and up-to-date computational fluid dynamics benchmark than the wizened Euler3D. As it happens, includes a CFD test constructed with Microsoft’s HPC Pack, and the XiFoam solver.
More information on XiFoam. SPECwpc allows us to yoke every core and thread of our test systems for this benchmark.SPECwpc NAMDThe SPECwpc benchmark also includes a Windows-ready implementation of NAMD., NAMD “is a parallel molecular dynamics code designed for high-performance simulation of large biomolecular systems. Based on, NAMD to hundreds of cores for typical simulations and for the largest simulations.” Our ambitions are considerably more modest, but NAMD seems an ideal benchmark for our many-core single-socket CPUs.
Digital audio workstation performanceAfter an extended hiatus, —DSP 2017 and VI 2017—return to make our CPUs sweat. The DSP benchmark tests the raw number of VST plugins a system can handle, while the complex VI project simulates a virtual instrument and sampling workload., who kindly provided us with the Kontakt licenses necessary to run the DAWBench VI project file. We greatly appreciate NI’s support—this benchmark would not have been possible without the help of the folks there. Be sure to check out their many fine digital audio products.A very special thanks also, who cut us a deal to assist us with our testing. RME’s hardware and software is legendary for its low latency and high quality, and the Babyface Pro has exemplified those virtues over the course of our time with it.We used as the platform for our tests.
To simulate a demanding workload, we tested each CPU with a 24-bit depth and 96-KHz sampling rate, and at two ASIO buffer depths: 96, the lowest our interface will allow at a 96 KHz sampling rate, and 128. In response to popular demand, we’re also testing two buffer depths at a sampling rate of 48 KHz: 64 and 128. We added VSTs or notes of polyphony to each session until we started hearing popping or other audio artifacts.Apologies for the lack of results at 96 KHz and a buffer depth of 96 here. Thanks to something in the chain of Reaper, Windows 10, and our ASIO driver, our many-core CPUs couldn’t run the 96-96 test at all—we got popping and crackling from the get-go. ConclusionsLet’s summarize the reams of data on the preceding pages using one of our infamous scatter plots. To more accurately represent each chip’s price-to-performance ratio, we used real-world pricing data from Newegg where it was available and manufacturers’ suggested prices where it wasn’t.For the straight-and-narrow stock-clocked system, the story of the Core i9-9980XE is a simple one. Where Intel CPUs were already superior to the competition, the i9-9980XE offers some nice performance improvements.
In applications that can light off Threadripper WX CPUs’ rocket boosters, the i9-9980XE’s under-the-hood refinement can’t overcome the Threadripper 2990WX’s sheer core-count advantage.Overclock the i9-9980XE, though, and the 18-core chip both extends its leads considerably and closes some of the gaps the 2990WX opens. Part of that performance comes courtesy of Intel’s decision to resume soldering the heat spreaders to the die on its refreshed Core X chips.
We couldn’t take our i9-7980XE past about 4.4 GHz without fighting thermal limits, but thanks in part to the reintroduction of solder TIM, we were able to push the i9-9980XE to an impressive 4.5 GHz on all 18 cores without calling the fire department—all with attainable, off-the-shelf cooling hardware. Hallelujah for that.Presuming the potential errors that overclocking might introduce are tolerable in an i9-9980XE system, those willing to deploy custom liquid cooling loops may find that they can take all 18 cores of this chip to the limits of the underlying silicon, not just what paste thermal interface material might allow. Our casual overclocking suggests those could be some exciting limits to probe. Folks who don’t want to push the limits of their i9-9980XEs may find it easier to cool their chips quietly, too.The only complaint some may harbor about Intel’s refreshed Core X chips is that the company didn’t see any cause to cut prices on its high-end chips this time around. The reason, I imagine, is that AMD and Intel are trying to sell high-end desktop buyers two different stories of performance as competition heats up at the top of the CPU heap.In the $1200-and-up range we’re concerned with today, AMD seems plenty willing to make chips that really rip in workloads where sheer core count dominates, like rendering and some varieties of scientific computing.
The tradeoff is that Threadripper WX chips can fall far behind in other tasks. That inconsistency leads to Threadripper WX chips’ lower-than-might-be-expected standings in our overall value chart, even as they excel in a few particular workloads.Intel, for its part, seems unwilling to create chips with any corner cases, even if that means it can’t offer as many raw cores and threads as AMD can for the dollar. Core X CPUs may not take home the gold in every workload, but they’ll never put owners in a position where they can’t run certain tasks (or subsets of tasks) acceptably well, either. The extra money Intel CPUs command per core, then, is essentially insurance that you won’t be left hanging if the idea of inconsistent performance in any potential workload bothers you.If you know that a Threadripper WX CPU benefits your work and don’t care about the cases where it might not, then those chips can still be screaming performance bargains. Every person needs different things from a PC, so check around and see how your work maps onto AMD and Intel’s high-end desktop platforms before making the leap.If you have less than $1000 to spend on a CPU, Intel’s latest Core X chips might not move the needle much (presuming you even need four channels of memory and gobs of CPU-powered PCIe lanes to begin with).
We already know that the Threadripper 2920X and Threadripper 2950X offer better performance than the Core i9-7900X in many heavy-duty workloads for less money, and 10 higher-clocked Skylake-X cores may not close that gap much in the case of the $889 i9-9820X or $989 i9-9900X. That’s before we consider the $650-ish price tag on the remaining stock of Threadripper 1950X chips, as that processor remains quite formidable in its own right.If the Core i9-9980XE’s time on our test bench is any indication, though, the reintroduction of solder TIM, the advantages of the 14-nm process, and the end of PCIe lane segmentation from the CPU should make all Core X CPUs more attractive to those that want or need what they offer. Intel’s revised 10-core chips will almost certainly overclock better than Threadripper parts, and they should still hold a slight edge in whatever heights of high-refresh-rate gaming a builder might foolishly try to scale on high-end desktop platforms. How much did Intel pay TR to be their official advertiser? I was sure I saw a SHILL watermark in the background. Maybe if turn my monitor to this side I can see it again.
I am so sure I saw it!How many different people from how many different parts of the world did various piece of this review and what blind man assembled it? Did you even read your own review?Where are the stats for the 7980XE? The stats that show the 9980XE is the exact same chip with solder and slightly higher clocks, but with the same insane price tag? Since it replaces the former, that is rather a crucial point to leave out.Let’s try writing unbiased reviews by testing these products against the competitor, the previous part it replaces and list the facts, just the facts and let the reader form their own conclusions?It’s called unbiased journalism. AKA known as something that is damn near dead these days.
Bro, I’m at times critical of some aspects of the review, but you’re just rambling in a mostly dead thread, hoping no one calls you out for your fake criticism.A buyer of either CPU should perform research, and if not they are both CPUs from Intel, with the same core counts, and a little faster on the clock speed, will make it clear to even a best buy shopper that one is a little faster.Both CPUs don’t need to be listed for TR to still be unbiased.What’s your real issue? The one you made a fuss about isn’t much of one.#license2shill. Hi Jeff, longtime reader here.
I noticed recently you were looking for new benchmark ideas. Since the DAWbench benchmarks seem to be starting to cause controversy amongst fans of a certain architecture and digital audio production is an area that has many keenly interested, I have a benchmark suggestion that you can run alongside your DAWbench as a “second opinion” audio performance perspective. It uses Propellerhead Reason which has a free Demo mode which I think is sufficient for you to run the benchmark. If not, you can probably drop Propellerhead a line and maybe they might send you a license.url.
The question isn’t whether people make music on their computers, it’s whether DAWBench is a representative measuring stick.Do you have any desire to use 2000+ voice polyphony in your compositions?For many kinds of serious music production, people won’t frequently use more than about two dozen instruments. I imagine few people use much more than 100 instruments, and that therefore the non-DSP tests are simply irrelevant since the slowest CPUs of today sail over that bar with ease.I’m not confident in the benchmark. Perhaps it was a good reflection of real-world performance issues in audio production when it was developed, back in 2005. I suspect that in real world use today modern CPUs from both vendors are not bottlenecked in the fashion DAWBench measures. What was the last time that you opened a DAW and wrote a song?
What was the last time you spoke to an audio producer? Can you name some heavy-hitting plugins without googling? None of that needs the computing power of a modern HEDT-tier CPU outside of outlandish synthetic loads (a.k.a benches which are designed to showcase the difference between CPUs). You would be allocating more of your budget and concerns towards recording equipment, sampler and software to make it all happen then the CPU itself.9700K and 8700K in the DAW suite is more then up to the task of real-world workloads without breaking the bank. The real kicker is finding a mobile CPU that can effortlessly handle that and more without venturing into DTR land. You seem to have solid understanding of the mechanics of audio production, but your real world experience doesn’t seem to reflect the high end of production.I’ve been struggling with high end dual Xeon systems and the like for many years with composition rigs and post production, each for different reasons. Yes DPC latency is an issue you always need to optimize, but that is only to make a system under a full work load even operate at all.
CPU overhead has always been an issue with extreme sessions.I am here reading this article, because the work I do every day often hits a processing limit. FFT processing has become ubiquitous in modern post, and the software has grown to use the newer available horsepower. Now that workflows include much more of this more powerful technology, in series, clock speeds are more important than ever.Please stop passionately asserting what you don’t really seem to know. Hi Krogoth, I’m a fan of yours. You’re clearly very knowledgeable about computing, and a stand-up fellow all around.
But in terms of pro audio your comments aren’t wise to the reality. Orchestral composition (film/TV soundtracks) has been plagued by hardware limitations for at least 20 years. Realism and fidelity are perpetual moving targets – just like in graphics. Today many professional orchestral composers still use an array of slave computers over Gigabit ethernet, in conjunction with beastly workstations. If practically nobody wants to use over 200 virtual instruments, then the DAWBench VI test isn’t even “of academic interest,” it’s just a red herring benchmark, like using glxgears to compare video cards.
Just as getting a bazillion FPS in glxgears doesn’t mean you’ll get better performance in a real game, getting 3000 voice polyphony in DAWBench VI doesn’t mean better performance when using a sensible number of instruments in real production.It’s easier for me to imagine the DAWBench DSP test actually correlating with real-world use, but I still wonder how representative that is as well. I don’t doubt what you say is true. I don’t see why music production would not make use of CPU horsepower. However, I am incapable of judging whether this specific benchmark is realistic and accurately represent the performance in real-life music-related tasks. I hope it does. To me, it’s irrelevant because (in that order) (a) it’s an outlier and (b) I don’t do music on my pc.Whatever the case, it seems horribly non-optimized for AMD processors. I am wondering whether the developers have a specific explanation for this.
Is it a specific instruction set, like AVX256? It is a very specific memory access mode? A specific way of hitting the cache?
Is it the use of intel-specific libraries? Someone could ask them I suppose.
“The tradeoff is that Threadripper WX chips can fall far behind in other tasks. That inconsistency leads to Threadripper WX chips’ lower-than-might-be-expected standings in our overall value chart, even as they excel in a few particular workloads.”Much of this is because you test in Windows & not Linux Jeff.Windows has TERRIBLE NUMA support. In Linux (which is very likely to be the operating system of choice for people actually buying these CPU’s), the 2990WX matches or beats the i9-7980XE in most scenarios.url. Linux is most certainly not the choice of most people buying the 2990X. It’s just not true.2.
I’m all about running some real AVX-512 workloads on Linux to see what the Embree with Blender can do under Clear Linux. I’m willing to be after that you’ll stop complaining about how Windows ruins the Cinebench score.3. I’d love to see an intelligent response from the usual AMD defense squad telling us all how wonderful having massively unbalanced NUMA configuration is while pretending that Skyake X sucks when it flat-out wins numerous benchmarks against 32-core Threadrippers b. Maybe they should have called it the Core i11-9980XE. It’s certainly cranked up to 11.This CPU is only like 15% more expensive than the 2990WX, but for the performance it’s almost just a flat-out great value. There’s nothing the Threadripper outright wins like the 9980XE outright wins DAW Bench. It’s comedy how poorly AMD shows, and it makes Avid seem almost sane for only certifying Intel CPUs.
More 5X the plugin capacity at 96kHz with 128 samples of latency. That’s like 1.333 milliseconds of latency, if my math doesn’t suck, basically taking the computer out of the equation entirely at an insanely high bitrate.That’s intended to be high praise, not a backhanded compliment. These HEDT platforms usually completely disregard value, but in some cases it’ll triple up on a Core i7-9900K and more-than-double the 9900K on occasion, even with its clock speed disadvantage.edit: I saw the Ryzen 5 2600 make an appearance. Hope to see a full review on that soon. It’s a consideration for a PC I’m speccing out for my nephew.
Good on intel for.almost. managing to muzzle “the segmentation goblins”. Clock rates and performance look solid. Prices look a wee bit high, but not unreasonably so if your field benefits from the particulars of intel’s architecture.I maintain that intel would be better off selling all its enthusiast chips on an OC-able platform with unofficial ECC support like AMD does. The people who need verified support will go for the real Xeons anyway.intel could bin the very best chips (in terms of perf/watt) under the Xeon brand and then allow people to buy the chips that don’t quite meet the cut as K/X chips at slightly reduced (compared to Xeon) prices on enthusiast platforms with an unlocked multiplier and.still. profit from the arrangement. Aside from AVX512, something to note about Intel CPUs is that they support an extremely fast PDEP and PEXT assembly instruction: executing in just 1-clock tick.
AMD Ryzen supports PDEP and PEXT too, but it takes 18-cycles to execute. Basically, its only there for compatibility.For most people, this means nothing. But for some high-performance programmers and some weird cases (ex: Stockfish uses PEXT to calculate where bishops can move), this may be useful.256-bit AVX and AVX512 support are the big stuff for i9-9900xe and other -X CPUs from Intel. Don’t confuse 1 guys amatuer experience with reality. Phoronix doesn’t even know how to buy server ram apparently.
If they had at least tried micron or kingston I wouldn’t be facepalming.Anyone with a clue knows you want b-die for Zen cores, and you can pick up b. NEMIX, the RAM maker (or brand or the shipper) seems to be the issue. Evidences are stated below.1.) Michael says on his own forum (url. I was giving a short answer in a comments discussion.
Obviously most shops won’t spring for overspecced PCs because “why not they’re cheap LOL.” But there are many industries and applications where time is money, and saving a billable hour or two a day or pushing out a few more units pays for that $1-2K within a week or less. I’ve been on projects where the billable rate was so high that it would have made sense to assign me a dedicated IT person to prevent downtime, because saving several billable hours a week would have covered their entire compensation. That’s not a “need for speed” application, but the same concept applies.One exception is government departments with “use it or lose it” budgets who often end up splurging at the end of the year just to spend. Maybe, but when you’re buying 50 of these at once, that’s a big difference in cost that needs to be approved and when one of the beancounters mentions that you can hire two more staff for that price, you suddenly find yourself doing two more inductions for new staff.Software licensing seems to be getting much lower these days, too. In enterprises where the really expensive software is mandatory, the eye-watering list prices are i.