The latest Ask GN brings us to episode #70. We’ve been running this series for a few years now, but the questions remain top-notch. For this past week, viewers asked about nVidia’s “Ampere” and “Turing” architectures – or the rumored ones, anyway – and what we know of the naming. For other core component questions, Raven Ridge received a quick note on out-of-box motherboard support and BIOS flashing.
Non-core questions pertained to cooling, like the “best” CLCs when normalizing for fans, or hybrid-cooled graphics VRM and VRAM temperatures. Mousepad engineering got something of an interesting sideshoot, for which we recruited engineers at Logitech for insight on mouse sensor interaction with surfaces.
More at the video below, or find our Patreon special here.
As part of our new and ongoing “Bench Theory” series, we are publishing a year’s worth of internal-only data that we’ve used to drive our 2018 GPU test methodology. We haven’t yet implemented the 2018 test suite, but will be soon. The goal of this series is to help viewers and readers understand what goes into test design, and we aim to underscore the level of accuracy that GN demands for its publication. Our first information dump focused on benchmark duration, addressing when it’s appropriate to use 30-second runs, 60-second runs, and more. As we stated in the first piece, we ask that any content creators leveraging this research in their own testing properly credit GamersNexus for its findings.
Today, we’re looking at standard deviation and run-to-run variance in tested games. Games on bench cycle regularly, so the purpose is less for game-specific standard deviation (something we’re now addressing at game launch) and more for an overall understanding of how games deviate run-to-run. This is why conducting multiple, shorter test passes (see: benchmark duration) is often preferable to conducting fewer, longer passes; after all, we are all bound by the laws of time.
Looking at statistical dispersion can help understand whether a game itself is accurate enough for hardware benchmarks. If a game is inaccurate or varies wildly from one run to the next, we have to look at whether that variance is driver-, hardware-, or software-related. If it’s just the game, we must then ask the philosophical question of whether it’s the game we’re testing, or if it’s the hardware we’re testing. Sometimes, testing a game that has highly variable performance can still be valuable – primarily if it’s a game people want to play, like PUBG, despite having questionable performance. Other times, the game should be tossed. If the goal is a hardware benchmark and a game is behaving in outlier fashion, and also largely unplayed, then it becomes suspect as a test platform.
We already have a dozen or so content pieces showing that delidding can improve thermal performance of Intel CPUs significantly, but we’ve always put the stock Intel IHS back in place. Today, we’re trying a $20 accessory – it’s a CNC-machined copper IHS from Rockit Cool, which purportedly increases surface area by 15% and smooths out points of contact. Intel’s stock IHS is a nickel-plated copper block, but is smaller in exposed surface area than the Rockit Cool alternative. The Intel IHS is also a non-flat surface – some coldplates are made concave to match the convex curvature of the Intel IHS (depending on your perspective of the heat spreader, granted), whereas the Rockit Cool solution is nearly perfectly flat. Most coolers have some slight conformity to mounting tension, flattening out coldplates atop a non-flat CPU IHS. For this reason and the increased surface (and contact) area, it was worth trying Rockit Cool’s solution.
At $14 to $20, this was worth trying. Today, we’re looking at if there’s any meaningful thermal improvement from a custom copper IHS for Intel CPUs, using an i7-8700K and Rockit Cool LGA115X heat spreader.
Our colleagues at Hardware Unboxed have posted a set of Intel thermal tests on the new 8th Generation CPUs, looking into thermal throttling and boost behavior under thermal load. This is similar to much of our work in case and cooler reviews, where we often demonstrate over-time plots of long burn-ins (30-120 minutes), useful for determining where a CPU may taper-off in clock speed. In Hardware Unboxed's testing, additional benchmarks were performed on the Intel stock cooler and two aftermarket coolers, used in a multitude of CPU benchmarks.
For our audience that has liked our thermal discussion pieces (which is most of them, these days), we think HUB's work is worth seeing. It's a good demonstration of where and when thermal throttling might occur on a CPU, and helps to address the question of less-than-ideal thermal conditions for CPU benchmarking.
Delidding the AMD R3 2200G wasn’t as clean as using pre-built tools for Intel CPUs, but we have a separate video that’ll show the delid process to expose the APU die. The new APUs use thermal paste, rather than AMD’s usual solder, which is likely a cost-saving measure for the low-end parts. We ran stock thermal tests on our 2200G using the included cooler and a 280mm X62 liquid cooler, then delidded it, applied Thermal Grizzly Conductonaut liquid metal, and ran the tests again. Today, we’re looking at that thermal test data to determine what sort of headroom we gain from the process.
Delidding the AMD R3 2200G is the same process as for the 2400G, and liquid metal application follows our same guidelines as for Intel CPUs. This isn’t something we recommend for the average user. As far as we’re aware, one of Der8auer’s delid kits does work for Raven Ridge, but we went the vise & razor route. This approach, as you might expect, is a bit riskier to the health of the APU. It wouldn’t be difficult to slide the knife too far and destroy a row of SMDs (surface-mount devices), so we’d advise not following our example unless willing to risk the investment.
Newegg’s sale of the new AMD Ryzen APUs, including the R3 2200G (that we’re reviewing now) and R5 2400G, posted the APUs above MSRP by roughly $20. The R5 2400G retailed on Newegg for $190, versus a $170 MSRP, and also landed the product significantly above Amazon’s competing pricing. We purchased APUs from both Newegg and Amazon, and paid less for the product from Amazon; of course, AMD (and other manufacturers) can’t control the prices of retailers – that’d actually be illegal – but they can certainly find ways to suggest a price. It is, after all, a manufacturer’s “suggested” retail price.
Today, we received the following note today from Newegg’s service account:
As we await arrival of our APUs today, we’ve seen a few news stories reporting “4.56GHz” overclocks (or similarly high clocks) on AMD’s new Ryzen+Vega amalgamation. Seemingly, this significantly higher overclock is achievable merely by entering S3 (sleep) in Windows, and is even easily validated with higher benchmark scores.
In reality, we believe this is the Windows timer bugging out, which has existed on previous platforms and CPUs. The bug is easy to replicate because it only requires entering S3 state – another commonly problematic Windows setting, based on a previous life of lab testing – and waking from S3 causes artificially high clock reports.
AMD's new Ryzen R5 2400G & R3 2200G APUs, codenamed "Raven Ridge," are available for sale ahead of official embargo lift. We've earmarked the pages, for anyone interested in getting a jump on the APUs. Note that, as always, we recommend waiting on reviews before purchase -- but we'll make it easier for you to find them. Our reviews of the 2200G & 2400G are pending arrival of parts, likely today/tomorrow, and we've pre-published some GT 1030 & low-end CPU testing. We'll fully finalize that content once the APUs are in.
For now, you can find the new APUs at these links:
Amazon R3 2200G listing (public - at time of posting, this was $99)
Amazon R5 2400G listing (private - will go live closer to 9AM EST)
Newegg R5 2400G listing (public - at time of posting, this was $190, a bit over MSRP)
Newegg R3 2200G listing (public - at time of posting, this was $130)
APU reviews have historically proven binary: Either it’s better to buy a dGPU and dirt-cheap CPU, or it’s actually a good deal. There is zero room for middle-ground in a market that’s targeting $150-$180 purchases. There’s no room to be wishy-washy, and no room for if/but/then arguments: It’s either better value than a dGPU + CPU, or it’s not worthwhile.
Preceding our impending Raven Ridge 2400G benchmarks, we decided to test the G4560 and R3 1200 with the best GPU money can buy – because it’s literally the only GPU you can buy right now. That’d be the GT 1030. Coupled with the G4560 (~$72), we land at ~$160 for both parts, depending on the momentary fluctuations of retailers. With the R3 1200, we land at about $180 for both. The 2400G is priced at $170, or thereabouts, and lands between the two.
(Note: The 2400G & 2200G appear to already be listed on retailers, despite the fact that, at time of writing, embargo is still on)
The PM02 is the successor to Silverstone’s Primera 01, a case we’ve often referenced but have never fully reviewed. As Silverstone points out on their site, Primera is Spanish for “first,” so please take a quiet moment to appreciate the name “Primera 02.” This (still) isn’t a review of the PM01, but since it will continue to be sold alongside the PM02 (~$140), we took this opportunity to do some testing and make a close comparison. The original Primera should be a familiar sight to anyone that’s seen our render rig or the community-funded gift PC from a few months ago. Because the two cases seen in those videos are being used, we acquired yet another for this article--we like the 01 a lot.
This review will benchmark the SilverStone PM01 vs. the PM02 and RL06 airflow PCs, with additional testing conducted across other popular on-market cases, including the H500P.
We moderate comments on a ~24~48 hour cycle. There will be some delay after submitting a comment.