- Joined
- May 2, 2017
- Messages
- 7,762 (3.04/day)
- Location
- Back in Norway
System Name | Hotbox |
---|---|
Processor | AMD Ryzen 7 5800X, 110/95/110, PBO +150Mhz, CO -7,-7,-20(x6), |
Motherboard | ASRock Phantom Gaming B550 ITX/ax |
Cooling | LOBO + Laing DDC 1T Plus PWM + Corsair XR5 280mm + 2x Arctic P14 |
Memory | 32GB G.Skill FlareX 3200c14 @3800c15 |
Video Card(s) | PowerColor Radeon 6900XT Liquid Devil Ultimate, UC@2250MHz max @~200W |
Storage | 2TB Adata SX8200 Pro |
Display(s) | Dell U2711 main, AOC 24P2C secondary |
Case | SSUPD Meshlicious |
Audio Device(s) | Optoma Nuforce μDAC 3 |
Power Supply | Corsair SF750 Platinum |
Mouse | Logitech G603 |
Keyboard | Keychron K3/Cooler Master MasterKeys Pro M w/DSA profile caps |
Software | Windows 10 Pro |
Looks like a pointed submission to me. Why did the dude run the 10900k with slower memory when he could've run it same same?
I would guess the person in question doesn't have access to sufficient hardware to do a like-for-like comparison. To be fair, 2933 is the highest supported memory speed on Comet Lake, so they're in line with AnandTech on that point.Also the memory config is different. Possibly just two ranks on the older system and definitely four on the new one.
To address the last point first: the ES chips I've seen circulating have all been desktop chips, but there might of course be some I've not seen. They're pretty huge chips - 8 large and 8 small cores (some chips have some of these disabled), which by itself makes it highly unlikely that these are destined for a mobile use case. Intel have been struggling to get Tiger Lake-H out the door, so even if the additional area from the 8 small cores isn't all that much, moving past that to an even larger die seems unlikely IMO.I didn't say there were no power consumption tests besides AVX512, and I never said you've made up any numbers. I know, because I read the article before I, you know, started this thread.
I said consumers, what matters to consumers. AVX, AVX512 and Photoscan are used in the three tests, not really consumer friendly to me. Besides that, only the freaky AVX512 result ended up in a comparing chart. TPU on the other hand shows power usage while gaming, which I find highly relevant, but that didn't happen here.
Now before you say "so what, we've got all info we need from those three power measurements". I know, most of us here thinks so, but it may not be clear enough for everyone.
All I'm saying is that it makes the article look bad, and the only power chart with competing CPU's happens to be that one exception so far that shows a crazy high power consumption.
It wouldn't be that much trouble to add gaming power consumption, and then have all four results (Photoscan, gaming, AVX, AVX512) in comparing charts.
Right now it's easy for some readers to think that RL is much hotter than CL, which I really doubt it is in most scenarios.
That's why it should've been nice to have a deeper dive into power consumption, because 290 W was a 8C WR, more or less. But like I said, I guess we have to wait a few weeks for that.
There's been ES CPU's circulating both with and without Atom cores, and AFAIK, not all of them are desktop.
As for the rest here: I think you're approaching the review from the wrong angle. AnandTech doesn't do general consumer oriented coverage, they do in-depth highly technical coverage presented in an impressively approachable way. They always present the reasoning in detail, and argue for their choice of benchmarks in depth. They obviously need to balance the various readerships they target, and tend to skew more business/enterprise/IT than most publications in their inclusions of office benchmarks, SPEC, and various workstation benchmarks. However, when they do so they also mostly argue how these compare to typical consumer workloads. Their #CPUOVERLOAD introduction article goes into pretty extreme detail on the reasoning behind their benchmarks.
As for the specific benchmarks chosen: they're not directly relevant to most consumers, but they cover the (currently three) major test points of all-core non-AVX, AVX2 and AVX512, and are thus reasonably representative for any all-core load in each of these classes. I understand the desire for a gaming power draw measurement, but as seen in TPU reviews, gaming power consumption testing on CPUs is pretty much meaningless. The 40W delta seen there between the lowest and highest CPU tested tells us that gaming is not a sufficiently intensive (or even) CPU workload to make out meaningful differences, especially when contrasted against the 210W delta in all-core loads. Most games stress a couple of threads heavily and a few more with light loads after all, and nobody should be using that as the basis for their cooling or PSU selection. It would certainly be interesting to see, for example, a graph showing CPU power vs. framerates across a few CPUs, but it's not a type of data that's of much use for anything. And if the argument is that games might scale to become more multi-threaded in the future, then the all-core load numbers are likely representative again.
Of course if what you're saying is just that only presenting a peak power graph visually, and no other visual comparisons, is a bit iffy, then yeah, I can agree there. It would be nice for graphs for each load category + the cumulative peak graph. But if the argument is that they should do so because idiots are going to share the peak load graph online and present it as if the Intel CPU consumes that much under normal loads? I don't think AT should base their editorial policies on the lowest common denominator of reading comprehension and the game of telephone that is idiots online screaming about their hobbies, so I don't agree wiht that.
Also, apparently I didn't spot that you were the OP here Guess that goes as a great example of "don't reply to forum posts when you're in a hurry". There really ought to be a blushing emoticon here.