- Status
- May 18, 2007
- 23,815
- 5,265
- 108,890
AMD being stuck at 28nm until Zen does not help that. At 14nm, AMD will have a whole lot more headroom to play with, enough so to have hinted at Zen APUs with HBM at some point in the future. With triple the memory bandwidth and up to quadruple the transistor budget to play with compared to DDR4 system memory and 28nm, the Zen HBM APUs should be quite interesting. Rumors say the first generation of HBM APUs will only use 128GB/s HBM (that would be single stack HBM1), which means we can expect performance somewhere in the neighborhood of the R7-370.
If AMD's Zen catches up with Intel on IPC and power efficiency, I might actually be tempted to get a Zen APU with the Polaris equivalent of the R7-370 for $200-250 instead of upgrading my HD5770.
Onus
Titan
- Jan 27, 2006
- 37,037
- 12
- 89,965
While I disagree with some of TX_Tech's observations, specifically the assertion of bias (addressed below), I agree with his observations regarding the merits of using a Skylake i3 in the comparison. Using one of Intel's latest IGPs would, IMO, make a lot of sense.
As to bias, I don't think it is in reviewers' interests (especially if they want review samples) to bash anything unless it is wholly ill-conceived, which is unlikely. A notable exception is utter junk, such as PSU-shaped objects, or other products possessing liar-labels, but that is not going to be the general case. This does not mean reviewers should lie; my own approach to a review is to ponder (and hopefully answer) "Who wants this, and why?" Companies do not invest their time, money, and effort to deliberately bring pointless products to market. You and your friends may not be in the target audience, and that's ok, but it doesn't mean the product is bad, or that saying another type of buyer should consider it constitutes bias.
As to bias, I don't think it is in reviewers' interests (especially if they want review samples) to bash anything unless it is wholly ill-conceived, which is unlikely. A notable exception is utter junk, such as PSU-shaped objects, or other products possessing liar-labels, but that is not going to be the general case. This does not mean reviewers should lie; my own approach to a review is to ponder (and hopefully answer) "Who wants this, and why?" Companies do not invest their time, money, and effort to deliberately bring pointless products to market. You and your friends may not be in the target audience, and that's ok, but it doesn't mean the product is bad, or that saying another type of buyer should consider it constitutes bias.
You are right if it's a kind of "best bang for the buck" article, but it is a single review of a product. it is nearly impossible to test all similar products in the same review and it is also impossible to buy each CPU. Intel doesn't sample this smaller CPUs due image problems 
Helio Filipe
Commendable
- Mar 24, 2016
- 2
- 0
- 1,510
- May 18, 2007
- 23,815
- 5,265
- 108,890
They are almost directly related: TDP tells you what the expected worst-case design power is supposed to be. The chip is unlikely to use that much power under typical loads but it should be possible to put together a workload that will push it that high. If you use a chip with half the TDP, you can reasonably expect it to use about half as much power on average.
The relationship under normal use is not 1:1 but it is close enough.
FormatC :
You are right if it's a kind of "best bang for the buck" article, but it is a single review of a product. it is nearly impossible to test all similar products in the same review and it is also impossible to buy each CPU. Intel doesn't sample this smaller CPUs due image problems
I understand "product review", but all too often (not just in this case) I see a product reviewed, then compared to other products that are not in the same class.
If you're going to provide gaming results, it is simply dishonest to the reader to not point out WHY the i3 is showing so poorly in the gaming results.
As for a sample, allow me to be blunt... if Tom's Hardware doesn't have a i3-6100 somewhere in the company, then someone isn't doing their job. Either get a sample from Intel, or simply buy one. That CPU is far too important to not have one, it is the key value/performance option from Intel under $200. Above that is the i5-6500 $200, below it is the G4400 @$65. Those are the three key chips to have in each comparison (add a i5-6600k overclocked to 4.5GHz if you like as well, as that is the further up sweet spot beyond the i5-6500).
Actually, in most cases an $80 video card is rather pointless these days. The lowest card worth buying, IMHO, is the GTX 750 TI, and even that is semi-marginal over built in graphics.
An i3-6100 on a H110 motherboard will be less than $200 total, use the stock Intel cooler, and you'll get within 75% of the performance of the AMD chip for less money, a cooler running system, and FAR more CPU power than AMD offers. If you ever want to game seriously, a GTX 960 is the perfect add in card to that i3 chip, but there is nothing you can do to the AMD chip to make up for the lousy per-core performance.
You could make an argument for the A10-7870K and a $50 motherboard, it does have slightly better GPU performance than the Intel chips, but you sure do end up with a lot of other limits to get there, IMHO. Amazon runs sales on that chip from time to time, I picked one up to test with a few months ago for $109, which was a fair price for that CPU. Today it is running $20 more than that.
AMD needs to pickup their game, as it stands, almost nothing they make is even worth considering, and that's sad.
Just for your interest:
I'm sitting in Germany, not in US, and bought only in 2015 hardware for over 3000 Euro by myself (and also built a complete audio lab, also paid from my own, private budget). I write not only about CPUs/APUs, so it's only a small part of my work. And take a look at my equipment (a few high-end scopes, clamps, infrared cameras, audio chamber, a lot of tools) and all the daily costs of a well-equiped lab (3 rooms, 5 working places with bench tables for consumer and workstation tests, electricity, taxes etc.). Some things are simply not in the prio list, this must be understandable and acceptable for the readers. Ok, in theory I'm able to buy all this things (mostly used for one or two reviews) if I reduce the costs on the other side. But I think it is better to pass some things and use the money better for longer living products as such entry-level CPUs and mainboards for one season.
FormatC :
Just for your interest:
I'm sitting in Germany, not in US, and bought only in 2015 hardware for over 3000 Euro by myself (and also built a complete audio lab, also paid from my own, private budget). I write not only about CPUs/APUs, so it's only a small part of my work. And take a look at my equipment (a few high-end scopes, clamps, infrared cameras, audio chamber, a lot of tools) and all the daily costs of a well-equiped lab (3 rooms, 5 working places with bench tables for consumer and workstation tests, electricity, taxes etc.). Some things are simply not in the prio list, this must be understandable and acceptable for the readers. Ok, in theory I'm able to buy all this things (mostly used for one or two reviews) if I reduce the costs on the other side. But I think it is better to pass some things and use the money better for longer living products as such entry-level CPUs and mainboards for one season.
Thanks for the reply and info...
I wasn't suggesting that you personally should buy one, I am saying that the company should buy one. For all the various tests to be done over the course of the year, and to continue follow up work in the years to come, it is one of those staples of a review bench.
If Tom's Hardware is expecting reviewers to buy stuff and doesn't have a location where they have a large collection of parts, well... that doesn't speak well of the site, to be frank. But that is probably beyond your control.
To be honest - you can't buy all. Each budget is limited, independend of the site or their philosophy. Don't forget - we are testing a lot of stuff, not only budget CPUs. You must be a billionaire to buy all this things for all Tom's locations worldwide. I asked Intel so much times, no response for this small CPUs . If a company like Intel isn't able to sample it in Germany - I'm not able to buy all their CPUs by myself. And secondary - I have access to a lot of products as loan project in a good store in the near. But this shop is waiting for this smaller CPUs since weeks - no shippings. You can't buy each part in each country worldwide if you need it urgent.
Onus
Titan
- Jan 27, 2006
- 37,037
- 12
- 89,965
TX_Tech, once again I agree with many of your points, but not all.
I certainly cannot speak for Purch, nor for its employees, but as a freelance contributor who does it because it is a hobby I enjoy, my parts can be put into one of four groups. Group #1 is a set of base components, provided by Tom's, that all reviewers have in order to put us all on level ground, making our results comparable. The PSU, RAM, system SSD, GPU, CPU, and cooler were all provided to me, since I review motherboards, and only that changes. Other motherboard reviewers have the same parts (unless the socket is different, in which case the CPU will also be different). Group #2 is the remaining "lab" equipment and supplies I need to do basic testing and writeups, for which freelancers are individually responsible, without which we wouldn't be able to do our reviews. For me, that means a case for the test builds, a UPS, monitor, KB, mouse, Kill-a-Watt meter, and miscellaneous cables, plus an entire other system with Microsoft Office (for compatibility). Group #3 is the review samples. Most of these come from the vendors. Tom's buys a few. I actually bought one. Group #4 is stuff I buy because I've grown a wild hair about something that looks interesting, or worth testing, or might improve the way I do tests. Examples here include the entire AM1 system I built and tested, a set of identical SSDs I added to my test environment, a network switch I bought as part of a revision of my test methodology, and a slightly stronger graphics card that I thought might make certain differences more visible. The money for this is "off-budget;" it doesn't come from my regular job, but from what I get for my articles, or from other side jobs.
I certainly cannot speak for Purch, nor for its employees, but as a freelance contributor who does it because it is a hobby I enjoy, my parts can be put into one of four groups. Group #1 is a set of base components, provided by Tom's, that all reviewers have in order to put us all on level ground, making our results comparable. The PSU, RAM, system SSD, GPU, CPU, and cooler were all provided to me, since I review motherboards, and only that changes. Other motherboard reviewers have the same parts (unless the socket is different, in which case the CPU will also be different). Group #2 is the remaining "lab" equipment and supplies I need to do basic testing and writeups, for which freelancers are individually responsible, without which we wouldn't be able to do our reviews. For me, that means a case for the test builds, a UPS, monitor, KB, mouse, Kill-a-Watt meter, and miscellaneous cables, plus an entire other system with Microsoft Office (for compatibility). Group #3 is the review samples. Most of these come from the vendors. Tom's buys a few. I actually bought one. Group #4 is stuff I buy because I've grown a wild hair about something that looks interesting, or worth testing, or might improve the way I do tests. Examples here include the entire AM1 system I built and tested, a set of identical SSDs I added to my test environment, a network switch I bought as part of a revision of my test methodology, and a slightly stronger graphics card that I thought might make certain differences more visible. The money for this is "off-budget;" it doesn't come from my regular job, but from what I get for my articles, or from other side jobs.
I do this in full-time for Tom's, but not as employee. Currently I take the money from this and a few other jobs (product development, quality management, consulting) to complete my lab and to expand a little bit. I'm trying to bring all my equipment to the perfection and very often are the last steps the most expensive investments
Here I was asked for the i3 6300, in Germany for smaller i5 (because hyperthreading with only 2 cores makes no sense for the future) and cheaper Pentiums (for the beginner). If I start to fullfill each personal request, nothing will left for scientific work except a whole bunch of (later useless) CPUs.
I plan a review for entry-level and lower mid-class computers after Computex with the new VGA cards (as How-To or DIY). Then you will see also your lovely i3's. But not earlier.
Here I was asked for the i3 6300, in Germany for smaller i5 (because hyperthreading with only 2 cores makes no sense for the future) and cheaper Pentiums (for the beginner). If I start to fullfill each personal request, nothing will left for scientific work except a whole bunch of (later useless) CPUs.
I plan a review for entry-level and lower mid-class computers after Computex with the new VGA cards (as How-To or DIY). Then you will see also your lovely i3's. But not earlier.
Takasis007
Honorable
- Feb 3, 2015
- 79
- 2
- 10,640
K so I am a little confused, why the i-7 6700 is even on a APU comparison.... A10 209$, i-7 515$...... Go figure right... If you want to compare APU's by all means compare them to the Intel chips that.... well... use on board graphics... For the price tag of the i-7 i could get the A10 and a GTX 960, I own 3 A10- 7850k's and they fit the bill perfectly for what they were intended. Why you would even think about doing CAD on one would be beyond me but what ever. With a budget, which these are intended for, they make perfect sense. I just recently had my disposable income where I could purchase my son his GTX 960 for his birthday and upgrade mine with a Radeon 270X. (a year after the APU purchase, kids are expensive) Please keep the apples with the apples and the oranges in their own baskets.... I really would however be interested in integrated graphics alongside other integrated graphics as a comparison.... That would be a good read.
I fully understand that you can't buy it all, but if you're a tech/PC review site and you're benchmarking CPUs/GPUs, there are some basics that you have to have. The i3-6100 is one of them. Putting a Haswell i3 in there against new chips without a word as to why the results are what they are is deceptive to the reader, either intentionally or unintentionally.
I repeat my first comments, the article as posted is biased to show Intel in a poor light in such a way that is simply not honest about the situation (this may have been unintentional, but it is there regardless). Comparing a $370 CPU against a $180 CPU is absurd, at least without a huge disclaimer on the subject or a posted reason for why. Further, putting a 2.5 year old CPU with a known poor iGPU up in gaming results against the Skylake i7 and A10 APUs also is deceptive. It creates the implication that the i3 is not good for gaming and that you have to spend $370 on Intel to get a decent chip, so go buy the AMD.
Not all people will read that into it, but many will, it is the top and bottom of the market, while ignoring the middle that belongs there.
If it was just a product review, fine, then talk just about it, don't put such comparisons up without actually posting what it is competing against. At the $180 price it is listed for, it really is against the i5-6400 (however the i5-6500 is clearly a better deal). And frankly if you're in the market for a A10, you shouldn't buy this until the price comes way down, since the A10-7870K is a far better deal, 200 MHz slower for 1/3 less money, otherwise the same chip. The new AMD cooler isn't remotely worth $50.
Final note: If the only two Intel chips that you have are the older Haswell i3 and the new Skylake i7, then frankly you aren't equipped to do CPU reviews, comparisons, and benchmarks and shouldn't have written this. You should have personal experience with all the current hardware on the market to even be able to write the article.
Example: Imagine a car reviewer for an auto magazine driving the new hatchback from brand X, and giving his review, yet he/she had never driven 2/3 of the cars on the market. That is what happened here.
---
Side note: You can actually simulate the i3-6100 pretty close by fixing the core speed to 3.7GHz then turning off 2 cores, while leaving hyperthreading turned on. I don't know of a way to disable 5MB of the L2 cache, perhaps there is, but that is one way to get close to the performance without physically having a i3-6100 in the office.
Takasis007
Honorable
- Feb 3, 2015
- 79
- 2
- 10,640
I think I may have goofed... And I apologize for it, is the i-7 in this review making use of integrated graphics? If it is, I sorely apologize for not staying up to date myself.. I still would love to see a side by side, like an APU hierchy chart like the Discrete GPUs have. To be perfectly honest, with the boys always wanting upgrades anything over a GTX 960 or R9- 380 just is not going to happen. Kind of curious about pairing the APU with a discrete as was hinted at with DX12 but haven't heard anything about if that was possible.
May be in North America. As I wrote - in Germany nobody is really interested in dual-cores for this price tag. SMT is only a "auxiliary crutch", but nothing real. If I know, that a review will be translated, I will try to satisfy also the US taste and organize such CPUs in the future. To be honest, nobody missed this CPU here in the original review...
And please: I hear always, that I'm Intel biased, now I'm AMD biased? Nice...
Gabriele Molinarolo
Honorable
- Oct 15, 2013
- 3
- 0
- 10,510
Btw the whole "3Ghz thingy while gaming" can be overcome with AMD msr tweaker. These AMD chips can be undervolted massively. So you can alter the p-states and set a much lower voltage than the predefined ones for each p-state.
By default, AMD sets 1.075v for the 3GHz p-state. I've been able to keep 3.5GHz with 1v on my 7870k. At 1.125v I can keep 4GHz. I use AMD MSR tweaker to keep all p-states locked to the same level, and consume less energy while going faster, compared to the default powertune settings.
By default, AMD sets 1.075v for the 3GHz p-state. I've been able to keep 3.5GHz with 1v on my 7870k. At 1.125v I can keep 4GHz. I use AMD MSR tweaker to keep all p-states locked to the same level, and consume less energy while going faster, compared to the default powertune settings.
Gabriele Molinarolo
Honorable
- Oct 15, 2013
- 3
- 0
- 10,510
megamanxtreme
Distinguished
- May 2, 2013
- 59
- 0
- 18,630
I explained it badly.
I was talking about a Zen APU, not CPU. (This article is solely based on the APU, unless it was a CPU review)
- Status
TRENDING THREADS
-
[SOLVED] Is it safe to use a recapped PSU for a long time ?- Started by zwtch_17
- Replies: 7
-
-
-
-
Discussion What's your favourite video game you've been playing?- Started by amdfangirl
- Replies: 3K
Facebook
Twitter
Instagram