upgrade, Intel or AMD?

[King of Snake]

I considering upgrading my computer soon-ish. For a long time the intel Core 2 Duo cpu's seemed like the obvious choice, but now I hear that the new AMD Phenom II processors have managed to catch up to Intel's lead and now offer very good value for money.
Anyone have any experience with these cpu's (and supporting motherboards) in combination with Scope?
I'm currently running and AMD Athlon 64 3200+ which has always served me well

[dawman]

AMD's lengthy experience with on board memory controllers has always given it an edge with VST FX counts.
I'd go for it, but don't get VIA, or AMD chipset logic. Go for NVidia.
And let use know how it works. I would love to see a Dual Phenom II Motherboard.

[kylie]

It is pretty much the same risk like going into a i7 Intel at this stage.

AndreD already posted positive results some time ago, and now I just found a review of a hamburg-audio pc in a german mag ( PC & Music) that seemed quite promising...

[valis]

The risks with Phenom & i7 aren't due to the cpu architecture themselves though, rather the board logic that the cpu's find themselves living in. Every cpu, chipset and board has errata, and experience has taught me over the years that at the very least it's best to wait for the 2nd or 3rd revision of a given motherboard. Waiting for revisions of chipsets & cpu's also pays dividends, but bugs here can at least be dealt with via BIOS updates & microcode updates delivered via the BIOS to devices on boot (like the cpu). However a bad board trace or choice of components somewhere on board will only be fixed via revision or RMA. There's no guarantee that they'll fix every issue that might affect you, just like there's no guarantee that even a version1 of a piece of hardware will even have a bug that affects you. It's all a bit of a lottery, which is where some will prefer to be 'out on the edge' and others may prefer to follow later once enough anecdotal info has built up from the community.

Personally I also like to wait until Intel gives the 'tock' of a given architecure. AMD doesn't conform to quite the same product cycle, but in practice that usually means you get a cooler running version of a given core (die shrink) that's been more optimized, has more room for enlarged cache(s) and potentially even a newer revision of the chipset(s) the cpu interfaces with. BUT, this is not the case this year, Intel is stepping over doing a 'tock' in the next die-shrink (32nm) directly to yet another core revision. So the best advice right now isn't to wait for the "tock" but just to research the specific board/cpu/component combinations as best as possible.

Nvidia's chipsets are subject to as much errata as the via/amd ones, the difference they is that they're pretty good about fixing things over time. So again a rev3/4 board of a given chipset is likely to have many of the earlier issues ironed out.

Also Phenom is quite competitive with Core2, but not with i7. i7 has a serious price premium at the current time though, and given that Intel is much worse about adhering to published specifications for later cpu's using the same socket I wouldn't assume a long upgrade path just because i7 is only debuting now. Many times voltage regulation and other considerations will invalidate a part for a given motherboard even though a year before it was claimed the upgrade path existed. AMD is *considerably* better in this regard, with only the most serious cases causing them to change engineering and removing backwards compatibility. In fact socket AM2, AM2+ and AM3 are quite cross-compatible, and there's no reason to believe that a new AM3 board wouldn't work for a decent lifespan.

[dawman]

Here's what my favorite hardware reviewer has noticed in recent build trends......

_______________________________________________________________________________

BIOS Ridiculousness: Everyone Say, "Thank You Gary"
Hey guys, Anand here. I'm writing this sub-section, not at Gary's request, but because I felt it was necessary. Over the past year I've watched the number of motherboards Gary gets to review go down, and the amount of time spent per motherboard go up tremendously. This year was especially bad as Gary spent more time helping manufacturers fix their BIOSes and compatibility problems than actually writing motherboard reviews.

I wanted to help bring some of what Gary does to light in this section, just so you know the sad state in which many of these motherboards are being brought to market and the work that goes into getting them ready so that we can actually write about it, much less recommend one.

With that said, let's take a look at a particular sequence of events we encountered with the motherboards in today's review. We are not going to name names today as all of the manufacturers are guilty, some worse than others. The point being is that we feel the lack of quality assurance before a product hits the market has now reached an all time high.

Testing Ridiculousness:
It is true, too true unfortunately, about the amount of time it takes to thoroughly test a motherboard, report problems, and then regression test a possible fix. I am anal retentive when it comes to this process as others are also. While the benefits of doing it eventually payoff for manufacturers and users alike, it is a disservice to our readership to delay reviews of new products based on this seemingly never ending cycle of test, report, test, report.

So we are refocusing our efforts in generating quick and to the point reviews in the motherboard section. The manufacturers are going to receive two rounds of the test and report process before we publish our reviews. After this, we will provide short updates about the product over its lifespan in the market. We are also instituting a new process where we will purchase select products at retail and review them as is. This means no conversations with the manufacturers or access to the designers and engineers that we currently enjoy. We will utilize the latest drivers, BIOS, and utilities on the website in the same way you do when purchasing a product.

Our plan is to cycle through each manufacturer so we are not singling out any one supplier but we are going to be brutally honest in our assessments in these particular focus reviews. Our hope is that it will spur the manufacturers to improve their internal QA processes and focus on product usability at launch instead of setting a world record in SuperPI.

That said, let's take a quick look at the number of problems we encountered up until this week with our four boards in today's review.

1. Of the course of the past 30 days we communicated problems, suggestions, and resolution status on our test products via email 896 times and over a 100 phone conversations.

2. We have received 31+ different BIOS releases in the last thirty days to address problems and/or improve performance.

3. Our change log of problems and fixes reads like a bad novel. While we will not ding the manufacturers for performance improvements that we or others suggest, the simple fact that auxiliary storage controllers, power management features, memory and voltage settings, and other basic features on these boards failed to even work or resulted in a non-POST situation just floored us.

We are talking about $300 plus motherboards designed and released to be the crown jewel in the manufacturers product lineup. Of course, there is no excuse for this regardless of price, but one would truly think that the QA process would have noticed simple items like S3 not working, drives attached to certain storage ports not recognized, 12GB memory configurations causing non-POST situations, various BIOS settings not working or auto settings generating out of bounds voltages at stock speeds, power management features that when enabled actually increased power consumption, various overclock bugs, and USB and network controllers operating at half speed. The list just goes on and on. To us, these were simple items that we found just booting the board and trying to use it in a manner that 99% of buyers would, not randomly generated bugs due to weird settings, bad drivers, or a collection of old peripherals.

Even more depressing was the fact that several of our technical contacts did not have the necessary components to recreate our problems in a timely manner. The biggest item was memory, specifically 12GB of memory. All of the boards had some type of problem with a 12GB installation, ranging from overclock performance to non-POST situations. Granted, 98% of the 12GB problems have been addressed now, but it took close to a month, dozens of phone calls, hundreds of messages, and constant pressure for this to happen.

We spoke with several personnel at various companies and they asked why we were so adamant about 12GB compatibility and performance (6GB operation was not much better at first) as one example. The typical response was not that many people will actually use 12GB and we tuned our board for high overclocks with 3GB, this is what the enthusiast wants. Our simple answer was and continues to be, "If you advertise the feature, we expect it to work correctly." This particular problem highlighted one area that seems to drive the current high-end market.

Catering or focusing exclusively to the extreme overclocking community has resulted in initial product launches that are focused on getting the highest possible results from a product at the expense of usability, compatibility, and stability. The quest to release quickly and have the top motherboard in the forums, or HWBot/FutureMark rankings has blinded some of the product teams to the more important issue of ensuring their product actually works as advertised. We enjoy seeing these records as much as anyone else and I am guilty of scouring the Internet everyday to see what record has been broken and more importantly, how it was done.

Overclocking is interesting to most of us and its importance in improving the quality of electrical components and design aspects on the motherboards cannot be overstated. However, we need balance in this area again. Simply, we need to get the basic features and options working right at product launch and then the BIOS engineers can have free reign in tuning the boards to reach their limits.

One of the technical marketing personnel at a particular board supplier kept pressing me on how well does the board overclock. They also wanted to know about 3GB memory performance at DDR3-2000+ and my SuperPI scores. I kept responding with a laundry list of items that needed to be fixed before I would even worry about overclocking. His responses continued to be, those problems are minor and we will get to them, what we need to know is if our board overclocks and performs better than the competition.

_______________________________________________________________________________

Intel has the best R & D and Quality Control for their products, but even so, they showed negative market growth which I am sure caused budget slashing, and since their competitors seem to release buggy hardware, it's safe to assume everyone else might follow suit.

I love reading on new dual i7 mobo's and get all excited at the new technology, but until I hear a certain DAW builder or 2 actually showing tests related to my app's I won't budge an inch.
But 2nd revisions mean price reductions too..............Sahweeet..

[valis]

Yes the lower clocked i7 is competitively priced against Intel's own c2q offerings, but the x58 boards are a price premium option too. DDR3 is dropping in price but yields, timings & clocks will jump when memory makers make the next process transition.

Really the main point to walk away with though is that these boards have made major changes on the Intel side which in theory should simplify things (removed the MCH aka northbridge) but as Jimmy points out not many manufacturers focus primarily on stability with the 'latest and greatest'. Even Intel occasionally has a stinker in their boards, and they make the reference themselves. However that doesn't mean you're guaranteed a smoldering pile of slag if you hook your Scope cards up, it just skews the numbers in the stability lottery in ways that won't be clear for our uses until more data is in on a given make/model.

Given the choice I'd go for Intel or Supermicro and skip the overclocking options from other makers anyday...(unless it's perhaps a dedicated gaming box in the corner of the studio?) Or of course go by the recommendations posted here. Others might be more adventurous, and I welcome them to have at it! We look forward to hearing reports of progress with new hardware for sure....

[valis]

The overall context of this discussion (imo) is that the music marketplace has shifted to having computers fully in focus now (DAW applications/sequencer/etc) and as such the community is 'strapped' to a shared set of computing values with the general public. Whereas 15-20 years ago a music workstation maker could pat themselves on the back for integrating hardware that might have been 5-6 years old for the general computing public (look ya'll get DISK DRIVES now!). Not anymore! (which is probably why many shifted their focus to cheap plastic devices that can be quickly mass produced by inexpensive labor forces...)

[Sounddesigner]

Given the choice I'd go for Intel or Supermicro and skip the overclocking options from other makers anyday...(unless it's perhaps a dedicated gaming box in the corner of the studio?) Or of course go by the recommendations posted here. Others might be more adventurous, and I welcome them to have at it! We look forward to hearing reports of progress with new hardware for sure....

You might not want intel mobo's with i7's. Several builders have stated that one of the x58 intel boards have problems with PCI/PCIE/dsp cards. I have the Gigabyte EX58 UD4P wich works fine with my SCOPE card and UD5 is working properly with Uad-2 so far. The Gigabyte x58 Mobo's so far are a success story with dsp cards and are the most recommended by Computer Builders. The Gigabyte X58 Mobo's are however reported to not work properly with MOTU 424 pci card by a builder. And i read on the uad forum that someone was having problems with the older uad-1's but that his uad-2 worked properly. With warranties, return policies, and using us early birds as ginnipigs there really is no risk.

The only problems i had is small things like my previous Asio module used would not work properly when i upgraded to a new system, i just needed to find a different asio module that did work with Sonar. There where other small problems but overall i have success with the Gigabyte x58 ud4p and SCOPE. The biggest test is Xite-1 ofcourse but for the last 2 months i've had no serious complaints with my SCOPE pci card running on the newer i7 Quads. I don't have to disable Cores and everything runs pretty smoothly, even with Hyper-Threading on.

@ Stardust, ADK computers has ran benchmarks that show DDR3 outperforming DDr2 when it comes to alot of samples at low-latency. I would assume that includes things like convolution IR's aswell. And i7's ddr3 1600 outperforms 1333 and other lesser speeds according to ADK.

[valis]

Thanks for the note in the Intel stuff, just reinforces the idea that for stability it's good to wait until a platform is tested & has some field reports in.

DDR3 is finally just beginning to show some gains but it's early in its lifespan, and yes right now there are certainly configurations that show performance improvements. But it seems to me that to get aggressive timings right now in many cases requires higher than the 'ideal' voltage as per the ddr3 spec and using the bare minimum of sticks. Another process shrink should bring chips to the market that run cooler with faster timings at better voltages, which will definitely push it farther in the market. Also DDR3 offers only the slightest performance increase when being paired with core2duo & core2quads as well, so purchasing it there might make sense if you think you're going to be building a new system right now, but I wouldn't look on it as a necessity.

Mind you I have built 2 systems recently for people using SuperMicro's C2SBX+ which are core2 & ddr3, so it doesn't mean I think you should run the other way screaming from that combination either. Just be realistic about what you expect from what you're buying. i7 & ddr3 are gaining maturity, but they're still not quite mainstream yet...

[Sounddesigner]

Also DDR3 offers only the slightest performance increase when being paired with core2duo & core2quads as well, so purchasing it there might make sense if you think you're going to be building a new system right now, but I wouldn't look on it as a necessity.

In ADK's benchmarks the difference between ddr2 and ddr3 for Core2Quads was huge. he used 2GB's memory worth of Samples with a 3.2GHz 1600FSB Quad with DDR3 and got 128 buffer vs
Quad core 3.0GHz 1600 FSB DDR2 wich got 512 buffer-size. 128bs vs 512bs is huge to many people and as jcschild pointed out in the thread (link below) the huge difference was not because of the little 200mhz speed difference of both quads but because of the ddr3 used.

Below is a link to the thread where jcschild (Scott from ADK) responded to and was talking about the advantages of ddr3 over ddr2 and how the price difference is now a moot point because the price amount of ddr3 is not that much more at all - http://www.gearslutz.com/board/music-co ... worth.html

Here's a link to the benchmarks - http://www.adkproaudio.com/benchmarks.cfm


In the end i agree with your arguement about waiting till the Nehalem platform and ddr3 matures further before buying i7 (If one can afford to wait), but also my interpretation is that ddr3 has made some significant improvement over ddr2 and the price difference is not big at all. But i agree on waiting till next generation 32nm Nehalem for a bit better performance cause i also don't get small buffer-sizes myself despite the current i7's low-latency performance advantages over other quads and ddr3 improvements . There are too many variables that keep most of us from getting and enjoying the low-latency performance seen in benchmarks. Tho some people will get that performance. This generation of i7's was a step in the right direction and still suitable for my needs but it is best to wait for the next generation if one can. but i would not buy Q9550/Q9650/Q9300/etc either, i'd either wait to buy 6-Core Westmere 32nm or buy this current generation of i7. Or unless i'm on a smaller budget and/or don't need extreme power/performance then maybe Q8200 with other cheap components, or a really cheap dual core. My opinions ofcourse.

[valis]

With C2Q & C2D. comparing ddr3 to ddr2 you only begin to see gains for ddr3 at 1333mhz, and it's not until 1600mhz that ddr3 *begins* to shine. Intel is only now shipping most C2D & C2Q parts at 1333mhz fsb (there are still some 1066mhz fsb parts in retail channels), and to get 1600mhz fsb stock you need the top Core2Extreme parts with a heck of a price premium, or you need to overclock. Since his benchmarks are running 3.2Ghz @ 1600mhz fsb I'm guessing he's overclocking 2.6Ghz/1333mhz cpu's or the Core2Extreme that goes for a rather hefty $1550 USD still (via newegg). I'm not at all surprised that ddr3 benchmarks will with overclocked (or overpriced) Core2Quad cpu's...but I personally prefer stability & reasonably priced parts.

Now, go look for 1600mhz ddr3 parts that meet the JEDEC spec of 1.5v or below. 1.575v is listed as 'maximum stable' voltage with 1.975v being the point to which parts must be able to absorb voltage without incurring damage (but note it is NOT specified that parts must WORK at this voltage.)

There are very few current ddr3 parts that operate at decent timings (cas latency refresh speeds) & overall operating speeds (mhz) without being overvolted. Most of the 1.5v ddr3 is listed as 1333mhz or below...this is what I mean by it being an immature process still. You CAN buy 1.5v parts that work at higher speeds, but have a gander at a retailer and see how many of these parts are listed versus 1.65 or even 1.7 to 1.8v parts. Even from Samsung, Kingston etc you'll find most of the faster ddr3 parts are typically 1.6 -> 1.65 volts.

Beyond this, the rest are chips binned by Samsung & other makers as being 'failed at 1.5v' and sold off in lots to 2nd tier vendors. This has been the game for the last few years. The 2nd tier ram 'packagers' (OCZ, Mushkin etc) take these parts and test them for the voltage will get them running stable and ship them out as 'performance parts'. And HEY bonus! these advanced voltages usually allow for advanced timings too, and so the parts fly out to overclockers who are used to overvolting anyway. The 'dark secret' here is the these makers accept the 25-50% return rate as the *norm*, as many of the parts will fail within a few days/weeks (and hey, if someone can try overclocking them further maybe we can just say it's their fault anyway.) It doesn't matter anyway because these companies got the 2nd tier binned parts at a really nice discount, even with the high return rate & loss from parts failure during testing they're still yielding a nice profit (or so I've been told).

So the point to take away there is that ddr3 is just starting to emerge in terms of its performance potential, it's still not a mature process but it's made gains versus 6-9 months ago when it began to come to market. Also there isn't any 1600mhz ddr2, there is 800mhz ddr2 run in dual channel though...

The next process shrink from the ddr3 chip manufacturers should allow most parts to operate at JEDEC spec voltages (or even below) and will bring lower temps & proper cas latencies to boot.

[Sounddesigner]

All technical has been said.

Why the hack should I search and pay for a rare 1600 DDR3 that fits my i7 which is overpriced as well, when the benchmarks showw that a 1066 DDR2 still does a good job at a much lower price.

Probably i am avaricious

I think we have different beliefs as to what's rare, and overpriced. One man's 'Bang for the buck' is anothers 'arm and a leg'.

A i7 920 costs $279.99, a mobo can be bought for $189, and 3gb tripple channel ddr3 can be bought for $54. I'm not convinced i'd spend too much less on a Q9550/Q9650/etc system but i am convinced that the performance gain of the i7 920 is VERY significant over the other processors. That significance can mean longer use and me not having to buy a new computer for a lot longer period of time, and better performance now. And the i7 does'nt cost much more but gives much more wich is different then most high-end products. I cannot attest to the low-latency performance of i7 (some people can) but i do see the difference with hyper-threading enabled and it is a VERY significant difference over not being enabled. Clock for clock i7 is faster then previouse generation quads and with hyper-threading it is far more powerful then them if you have a DAW like Sonar,Cubase, etc. That significance along with its price not being much higher then other Quads make it the best bang for the buck when going after great performance and power from my view. I'm pretty sure that the smaller the buffer-size used with i7 the more significant it's difference becomes over older generation quads aswell, wich means the more bang one is getting for their buck. I'm on Sonar 7 and the HT is a huge difference, i expect even better performance when i upgrade to version 8.

I agree with Valis on waiting till the i7 platform matures before buying if one is interested in it and can afford to wait, but if given the choice of i7 over last generation quads i'd choose i7 in a heartbeat, wich is what i did.


Memory - http://www.newegg.com/Product/Product.a ... 6820227376

Mobo - http://www.newegg.com/Product/Product.a ... 6813130221

Processor - http://www.newegg.com/Product/Product.a ... 6819115202

It's not hard for me to find i7 system parts and many companies are developing for the platform. The reason to go after i7 and ddr3 simply is better performance for dollars.

I don't know if my answers to your questions can be good enough for you so i'll end it with 'to each their own'.

[valis]

Nice to see you opted for 1.5v ddr3 parts from OCZ (but of course for the cas latency 7 & 1.5v it's a 1333mhz part.)

Stardust's take on "rare" may perhaps be due to not being US-based (or am I wrong?). Seems to me that UK/Eu costs are usually higher and supplies a bit less plentiful, especially as a product is in its first few months. I don't think i7/ddr3 is 'rare' but I would certainly agree that my take on the 'stability' and 'maturity' factor differs from some.

I try to target a platform for when it's on the "Tock" revision of a given architecture, and I try to keep an eye on Intel/AMD roadmaps (and other emerging tech) so that I am aware of when a good time to buy into new tech is (like waiting on the upcoming process shrink for ddr3). The debut of the "Tick" from Intel is usually when you have people posting up the first impressive benchmarks and overclocking numbers, and there's more excitement generated by the headlines about new tech. So it's a very attractive time to buy in, especially if one enjoys staying 'ahead of the curve'. And be sure if you need a new computer NOW, you pick what you can from what's available and make the best cost:performance choices you can. i7 offers compelling performance (wouldn't debate that at all), and prices are rock bottom in this economic situation we've got right now.

But I'm boring and look at my computers as tools where I appreciate stability & low-noise/heat output. Also I might add, in the last 5-7 years computers have passed the point where I constantly feel constrained by what cpu power I have available. Now certainly I can always utilize more if it's there, but I can also get by fine in the meantime by adjusting my workflow and never feel 'choked'. So I don't feel as compelled to upgrade constantly as I did in years back, at least if I avoid the constant upgrade cycles that Adobe/Steinberg/etc bring and the bloat that comes with most of them.

So for me, in i7's case it *would* have been another core revision this fall which is normally a decent time to buy in: debugged mobo revisions, cooler running cpu on smaller process etc. The problem this generation is that Intel has dropped the "tock", headed straight for another 'tick' with Westmere. Normally the jump would have been an i7 die revision at 32nm, but Intel is jumping the gun on both a new core & the process shrink at once. There's very little information on compatibility with current motherboards & etc for Westmere, one would hope that Intel retains compatibility but it just remains to be seen (and it's certainly going to depend on the specific build per motherboard). Of course manufacturers will state that they believe things will work fine, but i

It doesn't truly matter of course, even if many motherboards do require changes to their VRM's they'll hopefully have other bugs revisioned out and be using the same basic chipset/board layout. DDR3 will have rev'd a process itself and run cooler & faster...so yes if one can afford to wait I think the "Tocks" are a great time for the 'average' buyer to get into a platform. But I have no idea how that applies to things this year in all honesty!

[valis]

Also fwiw sounddesigner, techreport.com agrees with you:
highest-value offerings out of 16 popular Intel and AMD processors