Tom's 20th Anniversary: A Retrospective With The Editors-in-Chief

[EdJulio]

"DS: In 2005, we launched four new websites on a single day (TwitchGuru and other "Guru" sites that have been consolidated into the main Tom's site and the Tom's IT Pro site). That was crazy, but fun.? Barry Gerber was hired for that launch and has remained with the company ever since."

It's not Tom's IT Pro, it's Tom's Guide. We started Tom's IT Pro in 2011. Tom's Guide was created from all the consolidated non-TH sites way before then.

 

[jelyon]

It is crazy, exciting, and funny to think about the past 20 years when it comes to technology. In 1996 14.4Kbps was considered fast, for most people, and now quite a few people enjoy 1Gbps Google Fiber connections. Seeing the evolution of the optical drive from 2x to 8x to 32x was fairly fun, especially the ever increasing amount of noise the drives made. I have enjoyed reading the honest and in-depth reviews from Tom's Hardware over the past 18 years, and that honesty will be even more important in the coming twenty years.

Cybersecurity on smartphones is one area where vendors need to be called out when they either advance or diminish the security of a smartphone.
Testing the accuracy of health and wellness devices is also something that Tom's Hardware needs to do. Fitbit is already facing one class action lawsuit over the perception that its products are not accurate.
When the quantum revolution gets closer to the mainstream it will be important to have an honest insight into all parts of that revolution.

I am definitely looking forward to all technological advances, but those advances need the critical review from Tom's Hardware. Without an honest voice the coming twenty years will not be nearly as positive or fund.

Jesse Lyon

 

[SinxarKnights]

The best thing about Tom's is the removal of the huge blocking popup that asked what language you wanted every single time you came to the site.

 

[bit_user]

In 1996 14.4Kbps was considered fast,

Errr... 57.6k?

now quite a few people enjoy 1Gbps Google Fiber connections

I think it's still less common, as a % of US-based internet users. I definitely can't get it in my major metro area, and I had a cable modem in '98 (same town).

In some sense, the big story in tech isn't the relentless improvement, but rather the recent deceleration. I mean, why isn't 10 gigabit yet mainstream? Look at the wall that CPU (and to a lesser extent GPU) clock speeds hit? Of course, we all know that. But still, I think it's more interesting to look at the deceleration curve than to repeat the trite "gee whiz, back in 1996 I was using sticks to draw in the dirt!".

I mean, what are the realistic prospects & timetable for improvements in HDD capacity, CPU performance, memory capacity, network speeds, etc. And how does this compare to the last few decades?

those advances need the critical review from Tom's Hardware

I dunno. I could easily rattle off half a dozen others. Like, what about cloud services, or Raspberry Pi-class computers? How about more wearables coverage, in general? Drones, robotics, etc.

But Tom's (or the core site, anyway) has always seemed to focus on PC hardware enthusiasts, with a special focus on gaming. To some extent, they have to specialize. And it just makes sense to maintain focus on what's always been their mainstay. Tech is so pervasive and ubiquitous, these days.

I do think it's key to look at where the types of folks who were building PCs are now going. PCs won't be with us, forever. And Tom's will need to make certain transitions, in order to stay relevant, retain readership, and hopefully grow. I'm sure the know this, and I applaud their recent focus on VR.

 

[nukemaster]

Errr... 57.6k?

56k

I would say 28.8 and 33.6 would have been more common when Tom's was put online(but this depends greatly on your provider of the time).

 

[jelyon]

The pace of technological advancement is not slowing down, though. One of the biggest challenges today is writing software that is advanced as the hardware. How many games are written to take full advantage of an eight core processor? How many smartphone applications and games use three or more processor cores? Besides, if there is any slowdown it will go away when next generation lithium (or maybe even fuel cells) technology comes to the mainstream.

 

[turkey3_scratch]

The pace of technological advancement is not slowing down, though. One of the biggest challenges today is writing software that is advanced as the hardware. How many games are written to take full advantage of an eight core processor? How many smartphone applications and games use three or more processor cores? Besides, if there is any slowdown it will go away when next generation lithium (or maybe even fuel cells) technology comes to the mainstream.

Moore's Law is dead. I think it's going to be pretty hard to convince me that technological advancement is not slowing down. We're not seeing huge boosts anymore.

 

[jabliese]

Congratulations on the little ones, Tom! Good to hear from everybody.

 

[jimmysmitty]

Errr... 57.6k?

56k

I would say 28.8 and 33.6 would have been more common when Tom's was put online(but this depends greatly on your provider of the time).

Careful, you are scaring the kids. They don't understand what this old tech is.

The pace of technological advancement is not slowing down, though. One of the biggest challenges today is writing software that is advanced as the hardware. How many games are written to take full advantage of an eight core processor? How many smartphone applications and games use three or more processor cores? Besides, if there is any slowdown it will go away when next generation lithium (or maybe even fuel cells) technology comes to the mainstream.

Moore's Law is dead. I think it's going to be pretty hard to convince me that technological advancement is not slowing down. We're not seeing huge boosts anymore.

But it is not slowing down.

here is the thing, around the CPU everything is advancing very fast. CPUs have hit a wall but everything else seems to be going faster.

But at the same time in the enterprise market, CPUs are still getting big boosts. In the enterprise market, more cores is a good thing as is new instructions like AVX 2.0 etc. They use that, consumers do not.

In a modern server environment most companies use VMs to do their job instead of a server for everything. More more Cores they can fit per server the more VMs per server therefore the more productive a server can be and the less it costs overall.

 

[Onus]

It will be fascinating the day something like a networked render-farm, with hundreds of CPU cores coordinating tens of thousands of GPU cores becomes self-aware.

 

[kamhagh]

Woah, this website is 2-3 years older then me helped me a lot ^_^

 

[RedJaron]

It will be fascinating the day something like a networked render-farm, with hundreds of CPU cores coordinating tens of thousands of GPU cores becomes self-aware.

See, this is where it starts, and when we're all just organic batteries, guess who they'll blame? "This is all Joe's fault. What a tool he was. I have to spend all day computing pi because he plugged in the Overloard."

 

[Onus]

We'll all be saved from that fate by Turkey, who will deliberately hook up a Gutless Wonder to the Machine, which will fail at a critical time.

 

[jimmysmitty]

It will be fascinating the day something like a networked render-farm, with hundreds of CPU cores coordinating tens of thousands of GPU cores becomes self-aware.

See, this is where it starts, and when we're all just organic batteries, guess who they'll blame? "This is all Joe's fault. What a tool he was. I have to spend all day computing pi because he plugged in the Overloard."

All we need to do is divide by 0. Everyone knows that computers cannot divide by 0.

That or try to get it to say Aloha which means hello and goodbye thus rendering it in a never ending loop of confusion.

 

[Unolocogringo]

It will be fascinating the day something like a networked render-farm, with hundreds of CPU cores coordinating tens of thousands of GPU cores becomes self-aware.

See, this is where it starts, and when we're all just organic batteries, guess who they'll blame? "This is all Joe's fault. What a tool he was. I have to spend all day computing pi because he plugged in the Overloard."

All we need to do is divide by 0. Everyone knows that computers cannot divide by 0.

That or try to get it to say Aloha which means hello and goodbye thus rendering it in a never ending loop of confusion.

I like your style of thinking.

 

[turkey3_scratch]

We'll all be saved from that fate by Turkey, who will deliberately hook up a Gutless Wonder to the Machine, which will fail at a critical time.

I just got 3 in the mail

 

[RobAC]

Hey quit making me feel old!

Mr Pabst thank you for starting this site - it is fun watching this thing you create evolve from it's humble beginnings to what it is now. Looking forward to seeing your new creation. (And also to all the editors and contributors, thank you for giving us so much great information over the years.)

 

[Reynod]

A Great team of staff are behind the scene at Toms ... going all of the way back to the start. Always liked Thom's original articles and the highly polished stuff Chris produced, and continues to produce. Schmid and Roos articles have always been solid too ... for a couple of Intel fanbois (joke there guys) and of course I have to mention my buddy Crashman. Keep up the good work Fritz.

 

[bit_user]

Moore's Law is dead. I think it's going to be pretty hard to convince me that technological advancement is not slowing down. We're not seeing huge boosts anymore.

But it is not slowing down.

here is the thing, around the CPU everything is advancing very fast. CPUs have hit a wall but everything else seems to be going faster.

For a while, the new process nodes came on fast and furious. But with even those stretching out, I think you're wrong. It wouldn't be too hard for someone to plot aggregate compute performance of CPU over time, and see how far the curve is deviating from a nice exponential. I'd also be interested in compute divided by $ and by Watts.

The most telling story will be to do this with GPUs, since the concurrency in their workloads enables them to be much more scalable to high core counts.

I think this will clearly show that conventional semiconductor tech is decelerating, as it nears its limit. Yes, the improvements are still coming - especially in areas amenable to high parallelism - but we've already fallen off the curve we enjoyed for so long.

I'd like to see something a lot like this, updated for 2016: CPU, GPU and MIC Hardware Characteristics over Time

 

[jimmysmitty]

Moore's Law is dead. I think it's going to be pretty hard to convince me that technological advancement is not slowing down. We're not seeing huge boosts anymore.

But it is not slowing down.

here is the thing, around the CPU everything is advancing very fast. CPUs have hit a wall but everything else seems to be going faster.

For a while, the new process nodes came on fast and furious. But with even those stretching out, I think you're wrong. It wouldn't be too hard for someone to plot aggregate compute performance of CPU over time, and see how far the curve is deviating from a nice exponential. I'd also be interested in compute divided by $ and by Watts.

The most telling story will be to do this with GPUs, since the concurrency in their workloads enables them to be much more scalable to high core counts.

I think this will clearly show that conventional semiconductor tech is decelerating, as it nears its limit. Yes, the improvements are still coming - especially in areas amenable to high parallelism - but we've already fallen off the curve we enjoyed for so long.

I'd like to see something a lot like this, updated for 2016: CPU, GPU and MIC Hardware Characteristics over Time

There is a difference between technological advancement and CPU advancement. Everything around the CPU right now is going very fast. We went from 100MB/s as a standard to 600MB/s being easily affordable. Memory speeds have pretty much doubled and within a year or so we will be seeing NVDIMMs and 3D Xpoint plus new ideas for interconnects (Omni-Path/Coherent Fabric) that are vastly faster than current ones.

Then we have USB Type-C with Thunderbolt embedded in it which is currently at least 2x faster than USB 3.1.

I am sure it will hit a wall at some point but by then we might see CPUs hit a spike again, I would say when we find a new material to use.

 

[bit_user]

There is a difference between technological advancement and CPU advancement. Everything around the CPU right now is going very fast. We went from 100MB/s as a standard to 600MB/s being easily affordable. Memory speeds have pretty much doubled and within a year or so we will be seeing NVDIMMs and 3D Xpoint plus new ideas for interconnects (Omni-Path/Coherent Fabric) that are vastly faster than current ones.

Then we have USB Type-C with Thunderbolt embedded in it which is currently at least 2x faster than USB 3.1.

I am sure it will hit a wall at some point but by then we might see CPUs hit a spike again, I would say when we find a new material to use.

Um, 100 MB/sec vs. 600 MB/sec of what?

Anyway, in terms of system performance, I agree that HBM2/HMC in CPUs will be a big story in probably 2017/2018. NVDIMMs will boost storage performance even more (though it's not like storage has been a bottleneck, for a while, though I'm sure certain cloud & big data apps will benefit immensely). And software continues to evolve better ways of harnessing multi-core and GPUs. I was making these same arguments to someone, just a few months ago, about why I'm not too gloomy about tech.

Obviously, there will be CPU advancements in years and decades to come. We'll even see a few spikes, as fundamentally different technologies & manufacturing techniques come into use. The concern is just that the current era seems to be winding down, and I think the data supports that.

I was actually contemplating this, and I think the best way to see it would be to plot energy efficiency of computation, over time. The cost of each new node is initially high, and then drops, as yield improves and more capacity comes on stream. The link I posted above does include Peak GFLOPS per Watt.

You'll have to follow the link, since embedding it using the img tag didn't work (probably blocked by his site).

 

[jimmysmitty]

There is a difference between technological advancement and CPU advancement. Everything around the CPU right now is going very fast. We went from 100MB/s as a standard to 600MB/s being easily affordable. Memory speeds have pretty much doubled and within a year or so we will be seeing NVDIMMs and 3D Xpoint plus new ideas for interconnects (Omni-Path/Coherent Fabric) that are vastly faster than current ones.

Then we have USB Type-C with Thunderbolt embedded in it which is currently at least 2x faster than USB 3.1.

I am sure it will hit a wall at some point but by then we might see CPUs hit a spike again, I would say when we find a new material to use.

Um, 100 MB/sec vs. 600 MB/sec of what?

Anyway, in terms of system performance, I agree that HBM2/HMC in CPUs will be a big story in probably 2017/2018. NVDIMMs will boost storage performance even more (though it's not like storage has been a bottleneck, for a while, though I'm sure certain cloud & big data apps will benefit immensely). And software continues to evolve better ways of harnessing multi-core and GPUs. I was making these same arguments to someone, just a few months ago, about why I'm not too gloomy about tech.

Obviously, there will be CPU advancements in years and decades to come. We'll even see a few spikes, as fundamentally different technologies & manufacturing techniques come into use. The concern is just that the current era seems to be winding down, and I think the data supports that.

I was actually contemplating this, and I think the best way to see it would be to plot energy efficiency of computation, over time. The cost of each new node is initially high, and then drops, as yield improves and more capacity comes on stream. The link I posted above does include Peak GFLOPS per Watt.

You'll have to follow the link, since embedding it using the img tag didn't work (probably blocked by his site).

Sorry I meant storage speed. Was half asleep last night typing it.

The funny thing is that all the CPU advancements are benefiting servers. More cores, more memory and bandwidth is loved there and in HPC.