Worldwide PC shipments decline for 14th consecutive quarter

[HEXUS]

There were 61.7 million units shipped in the first quarter of 2018.

Read more.

[ETR316]

And, i'm not blaming, BUT i'm sure "Meltdown and Spectre" didn't help. Next year I think things may pick up a bit.

[shaithis]

Only going to get worse as the move to the cloud continues and companies realise they don't need to refresh fleets as much.

Current component prices aren't exactly helping though.

[LSG501]

I can never understand why pc shipments have declined when we've basically had miniscule performance improvements and increasing component prices (in some areas excessively) for the last 5+ years.... it's just baffling....


Hell my 10 year old amd x2 (ignoring the OS side of things) can manage just about everything a 'normal' user uses their pc for (office, web that sort of thing), it's only gamers, businesses (renewal cycles, workstations etc) and those who want the 'latest and greatest' that really need an up to date pc these days unless it fails.

Laptops are the areas thats had the most progress because intel (and amd) has focused on power usage more than all out performance.

[chinf]

My 10 year old Core 2 Duo laptop is now struggling noticeably after the performance hit of the Meltdown & Spectre patches, so I'm keen to see which Ryzen-based laptops will be good buys.

[3dcandy]

But I've noticed many are not getting laptops either just prefer to use phone and work pc if they need to and perhaps a tablet at home

[deepblue08]

The growth in CPU performance has been pretty small for the last 5 years, throw in high RAM and GPU prices and you got your explanation. Remember the days when Pentium 3 to Pentium 4 was a huge difference? That was just 1 generation! Nowadays, it's hard to tell the performance difference between a 4700k and 7700k without benchmarks to prove it.

[3dcandy]

The growth in CPU performance has been pretty small for the last 5 years, throw in high RAM and GPU prices and you got your explanation. Remember the days when Pentium 3 to Pentium 4 was a huge difference? That was just 1 generation! Nowadays, it's hard to tell the performance difference between a 4700k and 7700k without benchmarks to prove it.

True - went to hospital the other day and the nurses pc was running dodgy copy of WinXP lol

[DanceswithUnix]

The growth in CPU performance has been pretty small for the last 5 years, throw in high RAM and GPU prices and you got your explanation. Remember the days when Pentium 3 to Pentium 4 was a huge difference? That was just 1 generation! Nowadays, it's hard to tell the performance difference between a 4700k and 7700k without benchmarks to prove it.

Initial Pentium 4's were slower than the best Pentium 3, there are better examples but I know what you mean

Back in the '90s I used to only upgrade when I could buy a CPU that was at least twice as powerful as what I was running in key benchmarks. I am starting to wonder if there will *ever* be cpus that are usefully twice as fast as currently available CPUs.

I think at this point we are largely down to replacements for PCs that break for home users or go over the 3 or 4 year lease agreement for business models.

[afiretruck]

Initial Pentium 4's were slower than the best Pentium 3, there are better examples but I know what you mean

Back in the '90s I used to only upgrade when I could buy a CPU that was at least twice as powerful as what I was running in key benchmarks. I am starting to wonder if there will *ever* be cpus that are usefully twice as fast as currently available CPUs.

I think at this point we are largely down to replacements for PCs that break for home users or go over the 3 or 4 year lease agreement for business models.

I think that nowadays the issue is less of raw performance, and more what "possibly performance enhancing" features are in the newer CPUs compared to the older models. Things like AVX can make a noticeable positive performance impact in applications designed to use it, when the general processing speed is the same as that of the previous generation.

I look forward to when I can replace my E5-1620 (v1) with something generally twice as fast, but that won't be for a good half-decade at this rate (especially if/when Intel find their hardware Meltdown fixes slow things down!).

[Xlucine]

We've just had a 2x performance increase across a generation. Just over a year ago a 4C8T CPU around 4 GHz was respectable, nowadays you can get 8C16T at around 4 GHz for similar money. Laptops used to be chuffed with 2C4T, now 4C8T is common. Not the chipmakers fault if the benchmarks you're using don't use their chips properly

I look forward to when I can replace my E5-1620 (v1) with something generally twice as fast, but that won't be for a good half-decade at this rate (especially if/when Intel find their hardware Meltdown fixes slow things down!).

A 1700 provides twice the compute at half the TDP

[afiretruck]

We've just had a 2x performance increase across a generation. Just over a year ago a 4C8T CPU around 4 GHz was respectable, nowadays you can get 8C16T at around 4 GHz for similar money. Laptops used to be chuffed with 2C4T, now 4C8T is common. Not the chipmakers fault if the benchmarks you're using don't use their chips properly



A 1700 provides twice the compute at half the TDP

My point was that raw per-core performance just hasn't improved at the same rate. Most of the improvements we've seen, core count included, require software support to take advantage of these improvements. A per-core boost of +100% would be instantly noticeable with all software, regardless of how old it is/how it was designed.

Have you tried writing software to take advantage of multiple cores? It's not easy, and you definitely don't get 100% scaling.

So no, we definitely have not had a 2x performance increase across a generation. We've had an up-to 2x processing capacity increase across a generation.

[DanceswithUnix]

A per-core boost of +100% would be instantly noticeable with all software, regardless of how old it is/how it was designed.

I'm not convinced a 100% increase is possible. Well not at the high end, perhaps in laptops where perf per watt matters.

Thanks to semiconductor physics, it is hard to clock faster and switches to 7nm and smaller aren't likely to help. Clocks beyond 3GHz get you diminishing returns, and current desktop processors clock at 4GHz.
Thanks to instruction set architecture fundamentals, going beyond issuing 3 instructions per clock is into diminishing returns (2 if you care more about power usage) and current processors are already up to 4 or 5 issue width.

In the past, Moore's Law has meant twice the transistors every 18 months to find a use for and those process improvements also drove clock frequency up, but now those bottlenecks have moved so process shrinks don't help so much.

So you can't clock at 8GHz without doing much less work per clock, and you can't issue 10 instructions in parallel (because every 6th instruction is a jump so that means predicting two jumps per clock which would get messy).

[afiretruck]

I'm not convinced a 100% increase is possible.

One can dream...

[DanceswithUnix]

One can dream...

Our best chance would be if everyone played nice so we can turn the virus protection off, that would be worth several generations of performance

[OilSheikh]

I blame the bloody tablets. Poor man's PC