Haswell i7 Engineering Sample Pinned Up Against i7 3770K

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
[citation][nom]warezme[/nom]ugh, yet another generation I will likely skip since I don't give a rats posterior about the built in graphics. Who would have thought I would still be cruising along and relevant on an OCed Q6600 i7 at 3.6Gz and now old X58 SLI motherboard with triple channel memory. All I have had to upgrade the last several years are the video card or cards and switched to SSD's. Without competition Intel gets fat and lazy. The fact it is an engineering sample doesn't matter, the samples are rarely that far off from the production except for some microcode updates to better communicate with whatever chipset they are designed to fit in.[/citation]
X58 was LGA 1366 for the first i7s with triple channel RAM while the Q6600 was 65nm Core2 Quad LGA 775 (X38/X48).

Correction aside, I hear ya. I've been on the Q9300 for almost 5 years and will upgrade to Haswell IF final numbers improve. At the moment the 3770K should double my encode performance which is the only reason I would upgrade as my old quad core still chews through all my games and programs.
 
if you are interested in encoding you might want to wait for haswell.

quicksync is by far the fastest way to encode and has the best quality of all the hardware encoders.

haswell will come with a new version of quicksync, probably with considerably better performance than what ivy has
 
[citation][nom]neon neophyte[/nom]if you are interested in encoding you might want to wait for haswell.quicksync is by far the fastest way to encode and has the best quality of all the hardware encoders.haswell will come with a new version of quicksync, probably with considerably better performance than what ivy has[/citation]
Oooo, very good to know. I can't wait to see those benchmarks!
 


probably not even when you take overclocking into account

truth is, sandybridge is faster than ivybridge when both are overclocked as far as you can push them on ambient cooling solutions

i would pit my 2500k against anyones 3570k that is on similar cooling to me. i can get 5ghz out of my 2500k with a simple, cheap, air cooler. ivy is lucky to see 4.6ghz on a similar setup.
 
[citation][nom]payturr[/nom]It's not a big deal - they don't have to increase performance cause AMD is so behind, if AMD however caught up, I'm sure Intel would have something up their sleeve, but I wouldn't worry. This performance is still decent, especially since its not gonna be a MASSIVE architectural change like Penryn to Nehalem.[/citation]

Penryn to Nehalem was not that big of an artchitectural change. The changes were mostly in integrating the memory controller, minor architectural tweaks, and significantly changing the cache structure (that's not including stuff such as FPU changes, but if I had to guess, I'd bet that the FPUs are similar too). Intel hasn't had a massive architectural change since Netburst except for their very low end offerings such as Atom.
 


I too have heard great things about quicksync's performance, but not about its quality. I've read that it doesn't have quality comparable to many slower methods of encoding/transcoding.
 


Sandy at 5GHz (an uncommon overclock for an i5-2500K, I might add) is not really better than Ivy at 4.6GHz (a reasonable overclock for the i5-3570K). On average, Sandy may be a few percent faster, but there is no discernible real-world perfomrance difference in that and the opposite can happen very often, especially with cheap cooling where Ivy excels (getting higher and higher end cooling doesn't affect Ivy as much as it affects Sandy due to Ivy getting hot not because of insufficient cooling, but because of higher and higher end coolers not being able to totally counteract the paste issue beneath the IHS) a little better than Sandy in many cases.

Also, for the more brave with their chips, removing the IHS on an Ivy K edition CPU and getting rid of Intel's crap paste beneath the IHS actually lets it take off with considerable gains over Sandy around 20%.
 
i disagree, the ipc of ivybridge was largely marketing. i saw how it benchmarked comparatively to sandybridge, at 100mhz more. it was 2-3 percent most the time. the odd time was better.

in some cases, the ipc was worse.

5ghz on sandy is not uncommon, it largely depends on the motherboard.

if u want to race a 3570k against my 2500k ill take that up any day of the week.
 


nope. of all the hardware encoding methods the quality on quicksync is established as the best.

the only higher quality method is a software encoding, using strictly the cpu. this method is very slow and the difference in quality is negligible imo. i think the only people who should be doing software encodings are people making official releases.
 


ya, we all knew that one case where ivy performed better in super pi had to be a discrepancy. however, the numbers still arent much better than ivy.

let's just hope it overclocks better than ivy. good chance it will as what was holding ivy back on overclocking was the compound used between the die and the heatspreader. hopefully intel learned something from its mistake.
 
[citation][nom]warezme[/nom]Nice conspiracy theory but there is a problem with that..., that was the same reason no one ran out and bought Ivy Bridge..., there wasn't that much difference between it and Sandybridge, thus no reason to upgrade. That part hasn't changed.[/citation]
The problem with sites like THG is you have a heavily performance-biased crowd that craves every incremental improvement.

For the average non-gamer person or office machine though, even the slowest PC you can build today will be good enough for the next 5-7 years as long as the HSF stays on properly, the OS stays clean and the system has enough RAM to avoid getting bogged down by swapping.
 
[citation][nom]InvalidError[/nom]The problem with sites like THG is you have a heavily performance-biased crowd that craves every incremental improvement.[/citation]
Nothing wrong with that really, but they tend to overlook the other gains like efficiency and the platform overall.

And i mean, come on, if you have Sandy/Ivy, WHY would you want to spend money so soon? I'll upgrade to Haswell from Core 2, I really don't mind 10% upgrades over Ivy. As far as i'm concerned, i'll see around a 50% gain over Core 2 per core at a given clock speed, plus i'll also see a 1GHz jump in clock speed. That's almost double the performance in the last 5 years for a quad core part. And, the power consumption will be less, plus platform efficiency and feature improvements, etc.
 
They're letting AMD catch up in CPU performance. Remember when Microsoft bailed out Apple? AMD is better for me anyway. Their APUs offer a better price-performance ratio than anything intel has to offer. I can have my Skyrim and run my 3 virtual machines too, without spending money on a dedicated GPU. Using an i3 to run 3 virtual machines and a home OS? Laughable.
 
[citation][nom]neon neophyte[/nom]nope. of all the hardware encoding methods the quality on quicksync is established as the best.[/citation]

Funny. Everything I've read suggests exactly the opposite. QuickSync is fast, but you trade quality for that speed.
 
[citation][nom]bit_user[/nom]Hard to see how they can deliver more than a couple % improvement per clock, especially if you don't count the new AVX/FMA instructions. What they achieved with Sandy Bridge was pretty amazing and will not be repeated.[/citation]
Actually, Sandy Bridge was itself a minor 10-15% upgrade. The revolutionary jump came with Nehalem and the integrated memory controller it introduced. I'm still happy with my 4 years old 45nm Core i7-920...
 
FYI principal engineer is seriously up the ranks at Intel (and most places). That's basically 15+ year veteran. Any higher and you have to have patents and published research and whatnot (fellowship).

So that dude knows what he's talking about.
 
[citation][nom]neon neophyte[/nom]i disagree, the ipc of ivybridge was largely marketing. i saw how it benchmarked comparatively to sandybridge, at 100mhz more. it was 2-3 percent most the time. the odd time was better.in some cases, the ipc was worse.5ghz on sandy is not uncommon, it largely depends on the motherboard.if u want to race a 3570k against my 2500k ill take that up any day of the week.[/citation]

Only a few percent is plenty for covering a 400MHz difference between 4.6GHz and 5GHz. At that point, the difference in performance is negligible and that was my point.

Furthermore, 5GHz on the i5-2500K is uncommon. It's not very rare, but the i5-2500K generally tops out around 4.6GHz-4.8GHz, not around 5GHz. The i7-2600K and especially the i7-2700K are what tend to hit about 5GHz. Of course, this is all ignoring extreme cooling because that extreme cooling is generally not used even among overclocking enthusiasts.

Most people also tend to not have extreme motherboards and furthermore, the motherboard with Sandy and Ivy really doesn't make that big of a deal for CPU frequency. A decent motherboard generally goes almost as far as a top-end motherboard on the LGA 1155 socket platforms that support multiplier overclocking. Cooling is generally more important than the motherboard nowadays.
 
[citation][nom]neon neophyte[/nom]pretty much everything you just said is wrong[/citation]

None of what I said is wrong and all of it is easily verified here at Tom's, Anand, and other such sites and more as well as with personal experience.

 


I find it difficult to trust a site like that for the average based on that such a site is likely to attract above-average overclockers with more extreme cooling much more than the majority, especially since it goes against what I see overall elsewhere.
 


4.6GHz Ivy versus 5GHz Sandy. 4.6GHz+~10% is about equal to 5GHz. Ivy is on average about 3-7% faster than Sandy according to Tom's tests as can be verified in several Tom's articles on the i5-3570K and i7-3770K. 4.6* even 1.03 for 3% is about 4.75GHz. Even going by your minimal estimation of 2% gets about 4.7GHz equivalent to Sandy. The difference between the two in gaming performance is negligible even in the most CPU bound games.

The i5-2500K generally doesn't quite reach 5GHz without extreme cooling. That is much more commonly reached or breached by the slightly better-binned i7-2600K and especially by the i7-2700K. Even if it did, it would still not make a discernible difference because the numbers would still be too close. So, whether or not it does make it that far is not the point because these numbers are all already very close either way. At worst, I'd only be a few percent off even if I am wrong, so it's an argument in semantics anyway and the point that Sandy doesn't have any real advantage over Ivy remains true.

Extreme cooling most certainly is rare. By extreme, I mean stuff such as top-end water and air cooling and better. The vast majority of CPU overclocking purchases have coolers that are far less than ~$100 USD spent on cooling and are usually either low or mid-ranged coolers that tend to hit only around 4.2GHz-4.6GHz or thereabouts with a few exceptions above and below that range due to exceptionally good CPUs and such. This is very easy to verify just by comparing the number of purchases of lower end and mid-ranged coolers to the number of purchases of top-end coolers.

Extreme motherboards are also similarly less common than lower end and mid-ranged motherboards. It is also easily proven how the motherboard is not as big of a deal with the LGA 1155 socket CPUs as it is with many others because well over 4GHz overclocks are the norm regardless of the motherboard so long as its a chipset that supports multiplier overclocks and you have proper cooling. Does a top-notch motherboard make a difference? Undoubtedly. However, the difference is not huge nor even large these days. For example, looking at common overclocks for low-end boards such as the ASRock Pro4 Z77 and comparing them to overclocks of the same CPU and cooler or at least comparable cooling with a high end board such as a Gigabyte Sniper Z77 or Asus Maximus V Formula Z77 shows that the difference simply isn't great, especially with Ivy.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom