In our tests here (more in our benchmark database), AMD’s 3990X would get the crown over Intel’s dual socket offerings. The only thing really keeping me back from giving it is the same reason there was hesitation on the previous page: it doesn’t do enough to differentiate itself from AMD’s own 32-core CPU. Where AMD does win is in that ‘money is less of an issue scenario’, where using a single socket 64 core CPU can help consolidate systems, save power, and save money. Intel’s CPUs have a TDP of 205W each (more if you decide to use the turbo, which we did here), which totals 410W, while AMD maxed out at 280W in our tests. Technically Intel’s 2P has access to more PCIe lanes, but AMD’s PCIe lanes are PCIe 4.0, not PCIe 3.0, and with the right switch can power many more than Intel (if you’re saving 16k, then a switch is peanuts).
We acknowledge that our tests here aren’t in any way a comprehensive test of server level workloads, but for the user base that AMD is aiming for, we’d take the 64 core (or even the 32 core) in most circumstances over two Intel 28 core CPUs, and spend the extra money on memory, storage, or a couple of big fat GPUs.
Aside from the artificial maximum memory limitation – which AMD put in place to protect its own Epyc processors – the 3990X is simply a masterpiece. To be able to get 64 cores and 128 threads for a relatively mere $3990 is nothing short of stunning, and while few of us actually need a processor like that, the 3990X shines like the halo product that it is.
You know, I criticized the last anandtech ryzen review over bias, but I saw none of that here. It’s a good review that sticks to the facts!
There’s no question AMD is doing exceptional work with highly parallel tasks that don’t trigger NUMA bottlenecks. AMD is way out ahead for these types of workloads and way more cost effective than intel to boot. Yet I don’t feel they tried to hide the negatives and they acknowledged intel’s frequency advantages. My only critique is that they should also test under linux.
All this progress from AMD makes me wonder if typical software publishers will be more inclined to take advantage of massive CPU parallelism in the future. The gut reaction is “duh, yes they will”, but we also need to consider that a lot of the easy parallelism is running on the GPU, which trounces even AMD’s 128thread CPUs at parallelism. I’ve said it before, these make tons of sense for enterprise servers, but my gut feeling it will remain of marginal value for average desktop use cases. The economics & scalability still seems to favor GPUs for graphics and physical simulation.
Anyways, I’d still love to have a 64C/128T CPU just to see what I could do with it. Some workloads can’t use a GPU, last time code compilation was brought up as an example. I’d be very curious to see just how many threads are useful before reaching diminishing returns. 128 threads in parallel would expose a lot of disk & network I/O bottlenecks, heck I’m already hitting these bottlenecks today. Low cost 10gbps, where art thou?
“Low cost 10gbps, where art thou?”
https://www.youtube.com/watch?v=-LytcXun4hU
About the core count, there’s no use if the cores can’t be fed at proper speed, hence DDR4-3200 and PCIe 4.0.
Kochise,
Yeah, his videos comes up often when I search, but he’s got a different idea of affordable, haha.
Quoting linus tech tips…
“So if you want to try this out at home, these cards are actually available on ebay for like $200-$300 a pop and then you’ll pay about $60 for a 3 meter cable like this one.”
I want to be clear that I don’t just want to play with 10gbps connections, I actually need it to permanently replace my current network switches. I’d prefer to get genuine products from an authorized seller with a warranty. Ebay sellers are usually gray market with no warranty and scanning for the best deals on ebay can put you at a greater risk for counterfeits, sometimes the risk is acceptable, but meh. I could put that aside since beggars can’t be choosers. At the very minimum I’d need 2 10gbps nics plus 2 10gbps switches to upgrade because my computers and servers aren’t in the same room (and I have other 1gbps peripherals that I still need to connect too). Unfortunately 10gbps switches with ethernet ports haven’t come down in price. The passive SPF connections that are commonly found are limited to short ranges, long range options are available (including fiber), but the active transceivers aren’t cheap on top of the prices for everything else. The house is already wired for ethernet with 10gbps capable cable, and I’d hate to have to run new cables through walls and floors to reach the other side of the house. One last detail that may sound extraneous but is important to me is that due to limited free PCI lanes in my computer, I want a 4 lane PCIe3 card instead of an older 8 lane PCIe2 card that would incur a performance penalty with insufficient lanes. All the new cards have PCIe 3, they’re just not cheap.
I’ve been watching the market since 2018 at least, but the prices haven’t really been dropping and some have even gone up due to tariffs I believe. I’d be hard pressed to spend what it would take to do it right on top of other more pressing upgrades. Given that the prices aren’t budging, if I could find a good enough deal to save some money with used and/or chinese suppliers, I might give into the risk & compromises and try that, but it’s still a lot of money for something I was expecting to be much cheaper by now.
Sorry about the rant, I just thought I’d explain my situation, haha.
When you’re already ready to put $4000 in a 64 cores processor, I bet investing a bit more into a genuinely performant network for say, $600, isn’t that much to get the best out of your investment. I mean, Holy Molly, 1.21 gigowatts !!!!
Kochise,
I don’t think you realize how expensive this stuff is; $600 total for the LAN upgrade would be very reasonable! However the estimate is way off. The cheaper managed 10gbps ethernet switches tend to be in the $650-$1000 range and well known brands are around $2000. Keep in mind I’ll need two of these to upgrade my LAN and on top of this I still need at least a couple of 10gbps nics for computers.
A much cheaper solution that I had considered is just link aggregation. I already have extra gigabit nics laying around and my switches already support it. I’m not a fan of having to lay down extra copper and running more ethernet cables in parallel. Another caveat is that while linux is capable of bonding any two ethernet adapters, windows requires the drivers to specifically support it. But these are just gripes, I would still do it except I was hoping to have affordable 10gbps options by now, which is much better in every single way. Given that 10gbps has not gotten more affordable, maybe I need to re-evaluate link aggregation, it just sucks that it’s less versatile and a fraction of the bandwidth.
Ive been contemplating adding one of these mikrotik 10gb switches To my Home for a small 10gb network, mostly for between the server and my desktop – but ive simply not found The right excuse just yet 😉 – but i also have exsisting cat 6 cabling and with the sfp modules, i Can reuse that. Would proberly cost less then $300 for me.
https://mikrotik.com/product/crs305_1g_4s_in
https://www.youtube.com/watch?v=MDiiHN0MPdA
I can tell that a good 1 Gbps network is already so much comfortable than a legacy 100 Mbps one, I can just imagine what a 10 Gbps would provide when you have the need for. Just like the Linus video tells us, working “in the local cloud” could never have been so easy with such network performance.
Moving to 10G is awesome even when you don’t have the need. 🙂 The LAN is so much quicker.
damp,
Thanks for mentioning it. Personally I’m very hesitant to replace ethernet with SPI. Ethernet cables are so versatile (and cheap) compared to SPI cables & transceivers. Even if I were to commit to SPI switches and NICs for 10gbps, I still technically have a lot of other ethernet equipment that still needs to be hooked up so, so I would need an SPI/ethernet hybrid switch.
That mikrotik switch would be good for a 10gbps home lab. It looks like it’s intended to hook up 4 10gbps computers to each other with a 1gbps ethernet uplink for internet and/or the rest of the LAN. If I were just playing with 10gbps or only needed to hook up a few machines locally this would fit the bill. However as a backbone replacement for my LAN, it falls short of my requirements.
The reviewer brought up another point that worries me…
I have too little experience with SPF to draw upon, so I don’t know how big a problem this actually is. But the absolute last thing I want is to upgrade from standard eithernet to potentially vendor-locked SPF cables and cards.
Here’s an example…
https://community.arubanetworks.com/t5/Wired-Intelligent-Edge-Campus/Cisco-SFP-compatibility/m-p/388052#M2800
And the solution…
Seriously?!? No warranty if you use cables from a different vendor? WTF! This leaves me with a very strong preference for standard ethernet cables that I already have installed anyways.
Bah, I have no idea why I typed SPI so many times instead of SPF. I know the right word and I used it correctly later. My brain is slowly breaking down 🙁
The problem is the transceivers, and not necessarily the cables. Fiber is fiber. Get the correct correct fiber cable, plug, and go. Twinax DACs are trickier since the transciever and the cables are fused together. Lots of vendors like to play stupid games with transcievers, but blocking third party transcievers is illegal. FS.com has a range of DACs and transcievers which they certify to work with equipment from different vendors.
ServerTheHome (https://www.servethehome.com/) does reviews on things like this for home users, and the forums are full of people who run pro grade equipment at their houses.
Stupid games.
I won’t worry about this that much. Of course, I buy refurbed equipment for my house which doesn’t have a warranty these exact reasons. 🙂
Flatland_Spider,
I brought it up in the first place because I’ve heard about people having problems with it in the past. What is your source for saying “blocking third party transcievers is illegal”? I’m genuinely curious about any specific laws.
3rd party vendors like startech claim to provide cables (ie transceivers) that are compatible with specific vendors, and I believe them, but my question is how do they technically achieve this if they’re not authorized by the equipment vendor? Is this a case of 3rd party transceivers only managing to be compatible because they self-identify to the host using counterfeit identifiers to trick it into thinking it’s an OEM cable?
Obviously for my initial purchase I would buy things that are guarantied for compatibility, but I hate the idea that down the line I might want to aquire new equipment and have to buy new transceivers/cables that I already have because of stupid and intentional vendor restrictions.
If I absolutely had to go this route, then I’d do it because I didn’t have a choice. But if I have a chance to avoid vendor locking up front, then all of my experience tells me to run the other way or I’ll have only myself to blame down the line. I’ve been extremely pleased with ethernet for decades and it isn’t much more expensive per port after you factor in the cost of cables/transceivers.
Is there any specific reason you’d recommend SFP?
IMHO SFP makes a lot of sense for an array of rackmounted switches for companies with hundreds/thousands of users because passive SFP modules are very cheap and efficient. However for my use case I don’t need to hook up many rackmounted switches, just a few on different sides of the house. All I really want is to upgrade my existing 1gbps ethernet network to 10gbps.
Awesome! I’m looking at this now…
https://www.servethehome.com/netgear-gs110emx-review-a-managed-gs110mx-switch/
Checkout point #6: https://blog.startech.com/post/sfps-101-six-fundamentals-of-sfp-transceivers/
It’s the same reason HP can’t block third-party refilled ink or toner cartridges. They can whine about it, but they can’t do anything about it. HP went to court about this, and lost setting a legal precedent.
It’s a valid concern, and the situation used to be worse. It’s gotten much better in the last few years.
QSFP+is the newest standard, and I don’t think SFP+ commands the attention of the network vendors like it use to, which is why we’re seeing all of the third-party modules.
Most likely. 🙂
So do I. I haven’t made the jump for my home stuff because of vendors playing stupid games and cost. I’d like to have a 16-port 10G switch as my backbone, but I haven’t been willing to spend the money.
I think we’re almost there though. The new 2.5G and 5G equipment should allow the scale needed to bring the price down. Realtek is already making 1/2.5/5 chipsets for $40 NICs. 🙂
It’s easier to find switches with 10G SFP ports, and they’re cheaper then they’re RJ45 only versions since they made more of them. 10G has been used for uplink ports and backbone switches which typically means fiber or DACs.
Aside from that, mainly just to get fiber or DACs. Fiber is doesn’t degrade on longer runs the way copper does, and DACs reduce impedance because they are one thing without the copper to copper interface.
Most of the time I’m dealing with servers, and little things like interference, impedance, and distance add up.
At my house, I need to cover a 15×10 room, so SFP only makes sense for flexibility and experimentation.
Flatland_Spider
I read that already, but they don’t cite any specific laws.
Also to the best of my knowledge, it’s not illegal for printer manufacturers to use vendor locking technology. I think the cases that went to court were about printer manufacturers suing 3rd party clones that succeeded in breaking their vendor locks. At best the defendants won the right the break the locks if they were able to.
I haven’t heard of any laws that makes vendor locked SFP cables/devices illegal. And while I accept that counterfeit clones could be given the green light under case precedent, that doesn’t fundamentally help me when vendor locking restrictions crop up and I have to buy new cables that I wouldn’t need if there were no damn vendor locking in the first place. Yes, maybe I could buy a clone instead of OEM, but I really shouldn’t have to buy anything I already have!
I agree with that. Obviously SFP gives you more options, but I personally don’t own a large facility where fiber would be necessary, haha.
When you factor in cost of cable & transceivers, the prices aren’t that different.
This keeps coming up… It was designed for the datacenter. 10Gig isn’t needed for regular home users, it runs too hot, and the cables are too finicky for the average home user.
10G doesn’t make a 100/20Mbps Internet connection faster, and most people are connecting to their network via WiFi anyway.
Most 10G equipment and NICs run hot. They’re finally getting better, but they still need more power and better ventilation then multi-gig equipment.
10G is reaching the upper bounds of traditional, twisted pair copper cables, and it makes more sense with a twinax direct attach cable (DAC) or fiber. 10G can technically do 100m over cat6, cat6a, and cat7 twisted pair, but the performance degrades the farther away from the source the destination is. Staying under 10m with twisted pair is a rule of thumb that pops into my head.
Some people can use it, but they are a small slice of the market. Let’s call them Prosumers. 😀
Flatland_Spider,
Be that as it may, sometimes you need more than 1gbps to avoid bottlenecks regardless whether you are a residential user or not. Conversely many businesses can have more modest needs than residential users too. The need for 10gbps is obviously much more a function of what you are doing than business/residential status. Same goes for internet bandwidth and computer power. Many businesses simply don’t need much bandwidth or CPU for light web browsing or POS systems compared to residential users who might do a lot more video streaming and gaming. My point is, it doesn’t matter who you are, if 1gbps is insufficient then I don’t see a point in saying “10Gig isn’t needed for regular home users” since some home users can still find it useful for them and it’s not for us to tell them otherwise.
From a vendor perspective, it may not make sense to cater to the residential market because they aren’t willing to pay as much, but that’s different than saying there isn’t a need.
I concur that it takes more power. But as you said, newer cards are improving, also green ethernet switches & NICs support efficient idle states so the rate of energy consumption can be reduced during periods of inactivity. My current equipment supports this, and I would expect new equipment to support it as well (hopefully).
https://www.excitingip.com/2534/an-introduction-to-ieee-802-3az-energy-efficient-ethernet/
People are getting good results. Personally I’m not terribly concerned about this.
I’m fine with that label 🙂
Indeed. However, what we do at home is on the high side of the bell curve. 🙂
I would like 10G+ to run a storage network at my house. I can do it over 1G, but VMs over NFS or iSCSI work better over 10G+.
If a business has servers, they could probably benefit from a 10G+ server network.
Yes, and regular home users are most likely streaming over WiFi. Some gamers are probably gaming over WiFi as well.
Even if they are hardwired, 1G is still faster then most Internet connections. Especially uplinks. Some people are lucky to get 1G Internet. Most are not however.
I will fully admit 10G brings some enhancements to 1G, and I’m ready for 2.5/5G equipment to bring those enhancements to the masses. I don’t remember what they are off the top of my head, for the record.
I’ve been making sure I get switches equipped with Green Ports. 🙂
Flatland_Spider,
Some do and some don’t. Some people, particularly with fiber to the house, need a 1gbps lan speed just to make use of the bandwidth they have through their ISP. (I’m very jealous of those who have this option!) .
Gigabit ethernet offered consumers incredible boosts, but that was two decades ago and it’s become a bottleneck again. Even our WiFi standards are quickly evolving beyond the bandwidth limits of our wired 1gbps wired lans. Consumers using the new wifi standards will soon be leaving gigabit speeds in the dust. IMHO it’s time that wired LANs do the same at least for the high end consumer market.
https://www.extremetech.com/computing/184685-what-is-802-11ax-wifi-and-do-you-really-need-a-10gbps-connection-to-your-laptop
For the home market/mid range anything needing faster than 1G wired connectivity is not going to be wired. We’re soon getting 6Gbps wireless, which will bee much cheaper than 10G wired solution. Plus WiFi is much more convenient for the average home/geek user. No need to wire your home or office! 😉
10G/100G ethernet will not see the dramatic adoption and reduction in cost than 1Gbps ethernet did, mainly because it will be lower volume as only data center, or other very specific applications that simply don’t generate the same market volume. So the costs will remain high.
javiercero1,
You could be right that if the 10gbps ethernet prices don’t start to really come down, that >1gbps wifi will end up being cheaper. However, for better or worse the technical shortcomings for wifi will persist. The standard unlicensed frequencies are limited and can become crowded. Unlike wires, getting additional spectrum may not be feasible. Also the range/coverage gets worse when you step into the higher frequencies. With 802.11AC, I’ve seen people loose half the bandwidth one room over, that’s just the way it is. With mmwave spectrum, the amount of bandwidth is insane, yet you loose nearly all the signal through a few sheets of paper.
For smaller homes/apartments, most users are probably fine with a single access point, but otherwise more access points may be needed to get good coverage around a property, but therein lies the problem…how do you hook up all these access points effectively? Like it or not the best & most robust answer is still to run cables to each access point. Consumers who are extremely averse to wiring up the house will have to invest in more repeaters or powerline ethernet and accept the resulting tradeoff in performance. Repeating over wifi spectrum uses more bandwidth, introduces more lag, and may not be very useful in providing service if the repeater itself doesn’t have great connectivity to the base station. I have a relative who was adamant about going the all-wifi route to avoid wiring, but he ended up being disappointed in the results (500-600mbps in primary room, and ~100mbps at edge of house using 802.11AC). He bought a repeater without much improvement. Maybe another repeater would help more, who knows. YMMV.
Maybe it’s just me, but I’d rather run a cable and get reliable results than messing around with an array of wifi repeaters to get a robust signal across the house. I feel a small investment in cable is worth it just to save your self the aggravation of wifi performance variability in the long run.
Nah. Decent access points have plenty of bandwidth and coverage, and putting a wireless extender in a room is not a big deal. The average joe can now set up a wireless network covering a decent size home trivially.
For The vast majority of home and small business use wired Ethernet will be phasing out or for legacy purposes. With 5g and the new a WiFi you’re getting half the performance of 10Gbps Ethernet for orders of magnitude cheaper. Plus most devices connecting to those networks are not PCs or servers
I’ve already seen 12Gbps sustained wireless in the lab. Then 10Gbps Ethernet Value proposition for non datacenter, archival, infrastructure type applications will be very small.
javiercero1,
I’m sorry but you can’t “Nah” the limitations of wifi. These are real problems experienced by real users, Most users will want to at least be able to get the full bandwidth they are paying the ISP for. Also anyone with a NAS can benefit greatly from having at least 1gbps speeds.
A huge difference between ethernet and wifi speed ratings is that while gigabit ethernet has a lower speed rating than 802.11AC, ethernet performance is measured in actual bandwidth whereas wifi speeds of 1.7gbps, etc, are only theoretical bandwidths that you won’t actually achieve.
Here are reviews for some popular brand 802.11AC routers, what kind of performance do you think they get in cnet’s single story test space?
https://www.cnet.com/reviews/netgear-xr500-nighthawk-pro-gaming-router-review/2/
https://www.cnet.com/reviews/linksys-wrt32x-mu-mimo-gigabit-gaming-router-review/2/
https://www.cnet.com/reviews/d-link-exo-ac2600-mu-mimo-wi-fi-router-review/2/
5ghz channels:
Netgear
price range = $245-$300
7 ft = 772mbps
25 ft = 530mbps
50 ft = 224mbps
Linksys
price range = $128-$300
7 ft = 770mbps
25 ft = 370mbps
50 ft = 116mbps
Dlink
price range = $140-$167
7 ft = 927mbps
25 ft = 568mbps
50 ft = 283mbps
Now all of these are very good compared to previous gen wifi, however they’re still inferior to 1gbps ethernet and you can expect worse across a big multistory house. 802.11AX is the future, so we need to discuss it, but there’s still not many products or reviews for it.
https://www.cnet.com/news/how-fast-is-wi-fi-6-our-latest-speed-test-results/
If true, then 802.11ax should finally put wifi on par with 1gbps, at least in the same room. With signal attenuation it’s a good bet that range will still be no match for gigabit ethernet links across the house. Nevertheless if you want to say it is subjectively good enough for you and most users, that’s totally fine by me. My argument was never that most users need such high bandwidth, it is that for those of us in the prosumer crowd who do, 10gbps ethernet is very compelling. I just want it to become more accessible without spending a small fortune 🙂
This is where tri-band/triple radios come in handy. Two radios to handle the local clients, and one radio for backhaul. This solves the problem with repeaters.
Creating mesh wireless products has been a big push as of late. Google has their thing, Netgear has their Orbi equipment, and Asus has added mesh abilities for all of their newer stuff.
My Orbi mesh setup works well. The all have two A radios, one BGN radio, and a couple of 1G Ethernet ports. One A radio is backhaul between the APs, and the other two radios service clients. One is even acting as an Ethernet bridge for my Roku which produces way too much heat for it’s own good.
Flatland_Spider,
Bah, I have a comment “awaiting moderation” with points I don’t want to have to repeat. Hopefully it gets approved.
Triband radios aren’t going to work for 802.11ax, which already use all the bands collectively in order to provide 1gbps+ ethernet equivalent bandwidth at least at close range.
But let’s focus on 802.11ac. You may be looking at 300-600mbps speeds in an adjacent room using one of two 160mhz channels. The multiband repeater connects to both the primary 160mhz channel and becomes a new access point on the other 160mhz channel. The ‘third’ band refers to a new 2.4GHz access point, but this is for compatibility with 802.11G/N devices and doesn’t add backbone capacity.
So what are the advantages of this setup? Well, in ideal conditions the multi-band repeater should double the range with comparable bandwidths. Since the radio bands don’t overlap the radios should not step on each other.
However there are cons as well…one significant difference compared to running two separate access points with a wired ethernet back-haul: All the wifi traffic on the repeater network is repeated on the primary wifi network. So if you have active clients on both the repeater AP and the primary AP, then it creates a bottlneck on the primary AP since all users are sharing the primary AP’s bandwidth even though the APs are on different frequencies (unlike two independently wired APs).
Additionally since every single packet on the repeated network has to be transmitted on both of 802.11ac’s 160mhz bands, this causes double the congestion for nearby access points competing for the same bandwidth (ie APs on different floors, neighbors, etc).
There’s no doubt some consumers prefer everything to be wireless but they should be aware of the tradeoffs and when these can become inferior to wired networks. The message I’d like to end on is that we need to treat wifi as a shared resource, because that’s exactly what it is. The more bandwidth we demand from our own wifi devices (by aggregating more and more channels) the bigger the congestion problem becomes in areas of dense wifi traffic. Furthermore we should educate people that many of these consumer wireless repeaters/meshes with omnidirectional antennas running on unlicensed frequencies are actually consuming more of the shared bandwidth, the effects of which can cause more packet loss and deteriorate the overall WIFI experience for everyone.
I didn’t want to come off as a luddite, but IMHO we should stop trying to oversell wifi’s capabilities or undersell the benefits of wired cables for those willing to install it.
> Where AMD does win is in that ‘money is less of an issue scenario’,
you sure it’s not Intel in that sentence? considering how expensive their stuff is?
Intel is more expensive and less performant overall.
“Aside from the artificial maximum memory limitation – which AMD put in place to protect its own Epyc processors – the 3990X is simply a masterpiece. “
It probably isn’t an artificial limitation. Epyc supports twice as many memory channels, as well as registered memory support. The limit to Threadripper’s maximum is that there aren’t unregistered DIMMs large enough on the market.
Registerd DIMMs allow for something like 4x as much memory per slot/channel, allowing for much more ram. If you wanted to cram 2TB of memory into the 3990x (Which might be possible – Threadripper2 would usually work with ECC, but it’s ECC functionality wasn’t validated, therefore it didn’t claim support), you’re going to be spending $8,000 on memory, but if you’re spending that much on RAM, you’re probably definitely going to want to spend the extra $1k on an Epyc.