IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,785
- 136
My mistake on the earlier post, Core i3 features Hyperthreading but is different from Core i5 for the lack of Turbo Mode.
The Intel marketing guys have got to be just salivating over all the "feature combinations" they get to play with in their market segmentation model.
Only one thing is ever going to save us from this entropy hell that marketing is bringing to us, AMD needs to release BD and it needs to make a clean sweep on the performance crown and have a streamlined simplified segmentation model.
The Intel monopoly is starting to remind me a lot of the Microsoft monopoly, marketing management must have all gone to the same MBA school - just how many different versions of Vista did we really need?
Do we really need 4+ variations on nehalem features in combination with varying clockspeeds within each feature-set variation? Good grief Charlie Brown.
lol the best part would be a review with the cover story:- "i3 vs x3"
Power of 3 hahaha....
EDIT:- After reading the article i think that the i3 3.06Ghz will equal about E8500 in performance and i3 2.93Ghz will equal a E8400.
What AMD needs to to is to release x3's at 3.0Ghz - 3.3Ghz to fight with the i3's.
Power of 3 hahaha....
EDIT:- After reading the article i think that the i3 3.06Ghz will equal about E8500 in performance and i3 2.93Ghz will equal a E8400.
What AMD needs to to is to release x3's at 3.0Ghz - 3.3Ghz to fight with the i3's.
I think a 3ghz 730/740 should be out before the end of the year</wild speculation>
Will people be buying dual core Core i5s? Sure, you get Turbo mode, but what's the point when you can just plain overclock an i3 to the same speeds? It seems to me that Turbo is only useful in laptops where max TDP matters. In desktops you can just get a bigger heat sink and keep on going.
Of course, then the problem with the quad core i5s is that they don't have HT, which, according to the nehalem review, means they will be at least up to 20% slower than i7 clock for clock.
Will people be buying dual core Core i5s? Sure, you get Turbo mode, but what's the point when you can just plain overclock an i3 to the same speeds? It seems to me that Turbo is only useful in laptops where max TDP matters. In desktops you can just get a bigger heat sink and keep on going.
Of course, then the problem with the quad core i5s is that they don't have HT, which, according to the nehalem review, means they will be at least up to 20% slower than i7 clock for clock.
Denithor
Diamond Member
- Apr 11, 2004
- 6,300
- 23
- 81
In some cases HT can even hurt performance. When you have a workload that already completely saturates all four physical cores there is no need to add more overhead in order to schedule the work going into the cores.
Back in the P4 HT days we saw the same thing. HT often helped in cases where there were multiple light threads being processed - but if you hit 100% load it would actually slow down the work being done. Lots of games ran better with HT disabled.
Back in the P4 HT days we saw the same thing. HT often helped in cases where there were multiple light threads being processed - but if you hit 100% load it would actually slow down the work being done. Lots of games ran better with HT disabled.
HT hurting performance is such a rare occurence and it's such a tiny amount when it does. It's a non issue. I would never disable it on a processor that supported it. Worst case scenario you have 99% of your normal performance, best case 130%. More often than not, you see gains and not losses.
Ignoring specific benchmarks, HT allows a single core to do two things at once much more smoothly than otherwise possible. It's a very nice technology, and is imo the most important reason to use a Core i7 over a Core 2 Quad. Imagining a hypothetical situation, suppose you're playing a game that uses all four of your cores. While you play, something like antivirus or windows update decides to auto update, or some other automated process decides to kick in. HT on or off will be the difference between your game studdering or not. That's a real world gain that is noticeable and that's worth it to me.
Ignoring specific benchmarks, HT allows a single core to do two things at once much more smoothly than otherwise possible. It's a very nice technology, and is imo the most important reason to use a Core i7 over a Core 2 Quad. Imagining a hypothetical situation, suppose you're playing a game that uses all four of your cores. While you play, something like antivirus or windows update decides to auto update, or some other automated process decides to kick in. HT on or off will be the difference between your game studdering or not. That's a real world gain that is noticeable and that's worth it to me.
I agree with Denithor.
HT hurting performance is not a rare occurence.
And the Worst case scenario is not 99% of your normal performance at all.
Of cource choise is always a good thing, so it nice that Nehalem supports it.
Also is HT going to play a huge role with 8-core CPUs (2011) and beyond?
It's a rare occurrence that the performance is hurt by more than 1%. And yes HT will matter for 8 cores and beyond because applications are increasing in the number of threads they use just as fast as core count is increasing. Run two cpu intensive multi-threaded applications and you're already maxing a quad core. Do anything else and performance will suffer. For games that use 4 cores already, doing anything else simultaneously already hurts performance.
The thing people overlook is that the number of threads modern applications are using is increasing over time, it's not remaining static. If everything only used one thread, then an 8 core processor might be near untouchable in multi-tasking. However, you can run single programs that use 2-4 threads today, running more than one of those maxes out your processor. It really is a never ending struggle. As processors gain more and more engines to crunch code in parallel, new software will constantly be made to try and extract all of the crunching power the processor has to offer, which cuts down on multi-tasking capability.
Sorry but you said a different thing in your previous reply:
So you said:
1.HT hurting performance is such a rare occurence
2.When we have that rare occurence, the ammount that HT hurts performance is tiny
3.It's a non issue
4.You would never disable it on a processor that supported it
5.In the worst case (which case?, you mean the case of when this rare occurence happens?) we have 99% of our normal performance
6.Best case is 130%
7.More often than not, we see gains and not losses
Anyway like I said, I like that Nehalem supports HT, the only area that we disagree is on the percentages that you giving in your statements.
You mean like encoding a H264 movie while playing Crysis?
I don't know any game that has more than 5-10% difference between a dual and a quad core CPU (same core CPU) (the only exceptions that i know are UT3 & WiC and also here
the difference between a triple and a quad core CPU (same core CPU) is no more than 5%)
So in my perception there are no games that use more than 3 cores at 100%.
Today I would be happy with even a dual-core like 8400 but I guess our everyday use is completely different.
Cache size really made a difference between the P4 days with HT versus the situation with an i7. When you load two threads onto one physical core you effectively cut the available L1/L2 cache per thread on that core in half, as well as reduce the effective amount of L3$/thread in the case of i7's.
Having that large L3$ to begin with, versus not having it all in the case of P4's, really helps to minimize the negative performance impact of concurrent threads within the same physical core.
At any rate I liken the arguments surrounding "HT or no HT" as being similar in logic to the folks who refuse to use multi-channel ram (dual and triple) because doing so increases the bandwidth at the expense of also increasing latency.
For your latency sensitive apps (super pi) the extra bandwidth doesn't help but the increased latency causes performance degradation...whereas for your bandwidth sensitive apps the reduced latency of going single-channel actually hurts performance versus going dual or triple channel.
At the end of the day, as a consumer thinking you are making educated decisions on how to balance your hardware features (HT or no HT, single-channel or multi-channel, raid-0 or raid-1, etc) it is really all just a fruitless endeavor unless you personally take the time and effort to systematically study your own usage patterns and generate real-world data to support your notions and opinions on the topic. (in my case the bandwidth made nearly no difference in my apps of choice)
As with all things in life and computers, there are two serial operations that must occur and the first is impacted by latency and the second is impacted by bandwidth - figure out which matters to you and the data will set you free. Wait around for someone else to tell you what to think and you'll only have the right answer if you happen to use your computer in nearly the same way as the person giving you their free advice.
The main point I'm trying to get across is that the performance hit from enabled HT is so tiny and it's so rare that it is even noticeable makes it a non issue. Disabling HT because you want to increase performance doesn't really make sense to me. Having HT on can give you benefits that you can perceive with the naked eye without having to look at a benchmark. Turning HT off will pretty much never increase performance by an amount that is noticeable without a benchmark number or fps counter.
I'm discussing things from a multi-tasker perspective. Of course dual cores will be perfect solutions for people who want to run one task for quite some time. If you just play games or whatever and you're fine with that, you don't really need more parallel processing power.
I on the other hand, put an extremely high value into a system's ability to do multiple things at once without slowing down. I want to be able to be playing a game, but also get some work done if need be without compromises. I want to be able to be running various virus/spyware scans at the same time while playing a game. I want programs that have an automatic update setting to be able to automatically update without me even noticing that the PC is doing something other than the game I'm playing. I don't want to have to shut down torrents or encoding or whatever I'm doing if I get the craving to play some Team Fortress 2.
People severely downplay or misunderstand the benefits of a Quad Core over a Dual. They say things like if you don't do fifty things at once, you won't notice the difference. I'll give you a scenario that uses only two programs that you'll notice the difference on. Play any game that makes 100% use of two cores(UT3, Quake 4, Supreme Commander, etc..). Now pause it, alt tab, and do an anti virus scan. Go back to your game. It will have a reduced frame rate if you were using a dual core. If you were using a quad core then performance would be identical to what it was before the anti-virus was run. If it was a dual core with HT, it wouldn't be as smooth as the quad core, but the performance drop and studdering would be reduced(possibly significantly).
HT is not as good as real cores of course, but it does mimic the benefits of having more cores, just to a lesser degree. Think of it like rounding out the rough edges of multi-tasking. I can't stress enough how much I dislike things hitching up, and HT really helps alleviate that. As time goes on, programs that only used one core will use two(Same with two and four and more). A dual core user could run two separate one thread programs that each used 100% of a core without performance suffering on either program. If a new version of one of those programs comes out that makes use of two cores, that user's multi-tasking ability was just significantly hampered. That's why I say you can never have too many cores, and HT is a very nice way to improve a system's responsiveness.
The game Grand Theft Auto 4 has huge performance gains going from a dual core to a quad core.
In the end, it really comes down to the end user's habits of what they do on the computer. Some people do a ton of stuff at once and some don't. I would argue however, that if users had more multi-tasking ability, their pc usage habits would expand to fill that ability.
Leaving the review apart i was a bit saddened with the TDP of 73W that is a huge leap from 65W for the E8xxx, but then again it does not eat as much as E8400 lets see how true this TDP is in reality...
Nemesis 1
Lifer
- Dec 30, 2006
- 11,366
- 2
- 0
For crying out load. THIs is not same HT as the P4! People saying that this chip will be comparable to X3 is a joke . I have one and its better than PHII x4 . Rarely does the x4 beat it . RARELY @ same clock . Above 4ghz its not even close.
Intel ran into a problem with this chip. A problem that I would love to have. This chip is to GOOD. Its place in time! Its power usage! Its turbo mode . Witch means a great deal to consumers not enthusiast. The fact it does 4 threads at 30% and above most the time. To bad most coders are over paid and under educated. Because of this moment in time its the perfect enthusiast and productivity cpu and will remain so for at least 2 years until programmers get with the program! As for tirbo mode not counting at same lock . Who made ya rule maker. A 3.o6 is just that . 3.06 ghz. Its is a bonus that its turbos up to 3.6 or higher. Sure we forum guys can manual O/C . But were small in numbers. The consumer will buy the 2core in droves and it will be the cpu of choice for both enthusiast and consumer alike . This thing does it all better than well.
I'm gonna quote myself here from a past thread, which coincidentally was a post you regarding this same mentality you have that programmers make a willful personal choice when they show up to work at 8am to not make parallelized programs that day:
But that 73W includes the IGP & MC, as in the Northbridge. The NB power was always added to the system power before, now it's included in the CPU. Boards that use Clarkdale will NOT have a Northbridge chip.
I believe we have a net gain here, or power loss if you prefer.
evolucion8
Platinum Member
- Jun 17, 2005
- 2,867
- 3
- 81
Intel is not pushing performance in the midrange line of CPU's, like you stated, both runs almost identical in performance, and both costs the same, upgrading from a $183 processor from previous generation to the newest $183 processor makes no sense. Probably the story would be different if AMD was more competitive in all segments, for such low price, I would rather buy the Phenom II 940 and overclock the hell out of it, the same could be done with the i3 but is a more expensive platform.
There's some scenarios like in sinthetic benchmarks where a similar clocked i7 running at 3.20GHz can run 50% faster than a similar clocked Q9770. And the Nehalem architecture isn't any wider, which means that Hyper Threading maximizes the use of the CPU's resources like the Front Ends (In which its width is usually underutilized on Penryn/Conroe architecture), Execution engine etc. But for a normal desktop user, it means little since the only consumer applications which get some benefits of it is when doing 3D rendering like with Maya, or media encoding.
But Hyper Threading can reduce the bubbles in the execution engine if its implemented right.
The Pentium 4 had a very narrow pipeline, very small cache and in the end it was never designed to be a great multi core/SMT processor (Look at the Pentium D). Hyper Threading could bring up to 30% in performance to the Pentium 4 in the best case scenario which was pretty much a rare scenario.
Without Hyper Threading or Integrated memory controller, the i7 would be simply a heavily tweaked, souped up Penryn CPU, both architectures aren't much different.
A good example of this, WinRaR, it uses the 4 cores of my Q9650 but doesn't max it, Divx Encoder uses the 4 cores but doesn't max it, the Mass Effect game and RE5 benchmark uses the 4 cores but doesn't max it, which means two things;
1 - That the software hasn't catch up yet and can't load the CPU enough, probably a CPU architecture limitation (Penryn and Conroe have 4-issue Front Ends that are known to be underutilized) or a software limitation.
2 - That the software can't load enough the CPU's resources and bubbles will happens in the execution engine, which means that using Hyper Threading can increase the CPU usage more.
Anyway, essentially we don't disagree.
I said that the only point that we disagree is on the percentages that you giving in your statements.And this is logical, since our daily use is very different.
Like I said: i like that Nehalem has HT
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,785
- 136
Regarding Hyperthreading, did people forget about Replay on the Pentium 4 or not know about it? The people who bought the i7 were surprised that HT didn't have a performance impact while the HT on the P4 did.
http://www.anandtech.com/cpuch...owdoc.aspx?i=3570&p=10
Look at Lynnfield HT on vs off. The difference in games is mere 2-3%. Is it worth turning off HT for that 2-3%?
Of course, but CPUs aren't about THE CPU only isn't it? Especially on big platform level change like Core i7 went through. Same went through the Athlon 64, but turned out to be a fantastic CPU.
I mean, the integrated memory controller and the HT is there, and there will be no versions that won't feature the IMC, so why assume it?
Clarkdale vs. Core 2 differences are equal to
=
Bloomfield vs. Core 2 Quad
Anyone who assumes Bloomfield was worth it would think same for Clarkdale.
Same thing nowadays. It's like deciding against Q9650 or a Q9450. Why the hell would they buy Q9650 if they can use a Q9450 and OC it instead? The overclockers are extremely rare, and the most of them do a very mild overclock.
Plus, its a smart overclock that will watch chip temperature and power. Good HTPC processor.
Brand doesn't seem to be any less confusing with E7x00 series with only 3MB L2 but same clock speeds as the 65nm E6x00 series, or comparing the 2.93GHz E7500 to a 2.66GHz E8200, to site some examples. I think the new naming is more clear than the meaningless variations they have now which would perform same.
http://www.anandtech.com/cpuch...owdoc.aspx?i=3570&p=10
Look at Lynnfield HT on vs off. The difference in games is mere 2-3%. Is it worth turning off HT for that 2-3%?
Of course, but CPUs aren't about THE CPU only isn't it? Especially on big platform level change like Core i7 went through. Same went through the Athlon 64, but turned out to be a fantastic CPU.
I mean, the integrated memory controller and the HT is there, and there will be no versions that won't feature the IMC, so why assume it?
Clarkdale vs. Core 2 differences are equal to
=
Bloomfield vs. Core 2 Quad
Anyone who assumes Bloomfield was worth it would think same for Clarkdale.
Same thing nowadays. It's like deciding against Q9650 or a Q9450. Why the hell would they buy Q9650 if they can use a Q9450 and OC it instead? The overclockers are extremely rare, and the most of them do a very mild overclock.
Plus, its a smart overclock that will watch chip temperature and power. Good HTPC processor.
Brand doesn't seem to be any less confusing with E7x00 series with only 3MB L2 but same clock speeds as the 65nm E6x00 series, or comparing the 2.93GHz E7500 to a 2.66GHz E8200, to site some examples. I think the new naming is more clear than the meaningless variations they have now which would perform same.
If you had full manual control over the Turbo mode, then it would be very interesting to see what kind of overclcoks you can get on it while still maintaining stability and preventing overheating. If it is something that is just going to automatically disable the minute you touch any OC settings, then it doesn't seem that useful to people here.
I actually said "at least up to 20%". They can be up to 20% slower due to the lack of HT, and potentially even slower than that due to chipset, dual channel memory and whatever other differences there are between the platforms.
I'll bet that for any user if you average out all the situations HT decreases performance and all the situations it increases performance you will come out with a net gain.
It is a futile concept anyway, the performance gain in a few applications from turning HT off are much less than the gain from OCing just a tiny bit, and that still keeps you the HT benefits.
If you OC your CPU to the very limit then chances are you probably turned HT and turbo mode off anyway
Its a good feature, and since I will be stuck with AMD platforms for a while it is a feature I hope they copy.
It is a futile concept anyway, the performance gain in a few applications from turning HT off are much less than the gain from OCing just a tiny bit, and that still keeps you the HT benefits.
If you OC your CPU to the very limit then chances are you probably turned HT and turbo mode off anyway
Its a good feature, and since I will be stuck with AMD platforms for a while it is a feature I hope they copy.
I would like to add to my previous post that another reason why I said "at least" was that the 20% number we got was comparing quad cores. I imagine you will be more constrained with a dual core than with a quad, so you could potentially see even bigger gains.
TRENDING THREADS
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 9K
-
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 7K
-
Facebook
Twitter