Tuesday, July 1st 2008

Intel Nehalem Turbo-charges Radeon HD4850 Benchmark

Intel Nehalem Posts Impressive CPU Scores with 3D Benchmarks

The rather lucky Taiwanese team of Tom's Hardware got their hands on an Intel Bloomfield engineering sample that has a clock-speed of 2.93 GHz, running on a Intel X58 chipset based motherboard made by Foxconn called Renaissance to evaluate a Gainward Radeon HD4850 sample. System details are provided below.

Of course, the benchmark lacks the advantage NVIDIA PhysX gives to the CPU score in 3DMark Vantage, but for a CPU alone, it is a more than decent score. The system secured P7182 at default settings with a CPU score of 17966. In 3DMark06, it churned out 12786 3DMarks with a CPU score of 5183. In the Crysis CPU benchmark, scores of 33.70 and 18.29 were recorded at 1280x1024 resolution with no anti-aliasing.Source: Tom's Hardware
Add your own comment

55 Comments on Intel Nehalem Turbo-charges Radeon HD4850 Benchmark

#1
lemonadesoda
Nice results, but is it getting more punch out of the 4850? In fact, not really sure what this has to do with a 4850 at all. Could you put into the title and say the same thing?
Posted on Reply
#2
freaksavior
To infinity ... and beyond!
by: lemonadesoda
Nice results, but is it getting more punch out of the 4850? In fact, not really sure what this has to do with a 4850 at all. Could you put into the title and say the same thing?
i have to agree, what exactly does the point out?
Posted on Reply
#3
btarunr
Editor & Senior Moderator
by: lemonadesoda
Nice results, but is it getting more punch out of the 4850? In fact, not really sure what this has to do with a 4850 at all. Could you put into the title and say the same thing?
The fact that it's bringing out good CPU scores that affect this bench, and that this is perhaps one of the first published benchmarks of a video-card on a Nehalem derivative causes me to use that. So you get a glimpse of how Nehalem affects a HD4850 bench versus other benches using other CPU's that are all over the internet.

Another reason why 'HD4850' was used in the title was to show there's no latest NVIDIA card that could affect CPU score and that there are pure CPU scores in the benchmark.
Posted on Reply
#4
tvdang7
with me e6600 runnining 3.3ghz and my 4850 running 725/1025 my crysis bench runs at 31 frames. so that cpu does matter it beats me be 2 frames at default settings probably 4 or 5 frames overclocked.or more!!:banghead:
Posted on Reply
#5
DOM
well in 06 it gets 200 more points on the cpu vs mine at 3GHz :ohwell:
Posted on Reply
#6
Recoba
Better than my system: Q6600@3.6Ghz & ASUS EAH4850 score P7079 in Vantage. But why they run dual Channel ?
Posted on Reply
#8
Cold Storm
Battosai
I can't believe that they used VGA monitor for the bench! You'd think they would use the DVI feature?
Posted on Reply
#9
mullered07
by: Cold Storm
I can't believe that they used VGA monitor for the bench! You'd think they would use the DVI feature?
how will that affect the performance, you could argue there would be a negligable visual increase (dependant on your eyes lol) but nothing further :wtf:
Posted on Reply
#10
Cold Storm
Battosai
by: mullered07
how will that affect the performance, you could argue there would be a negligable visual increase (dependant on your eyes lol) but nothing further :wtf:
I was just saying that you would think of them as using a DVI type monitor instead of having to use the VGA extension.. That's all I was pointing out..
Posted on Reply
#11
farlex85
by: DanTheBanjoman
I'm not that impressed.

http://forums.techpowerup.com/showpost.php?p=823121&postcount=319

Sure I have 8 actual cores, though a lower clock, crap memory bandwidth and an inferior architecture.
Harpertowns at the same clock would perform a lot better. ie current generation can keep up easily.
We're still a ways away from it's release, and I believe these boards are still very new. It will likely speed up a decent amount before it's release.
Posted on Reply
#12
mullered07
by: Cold Storm
I was just saying that you would think of them as using a DVI type monitor instead of having to use the VGA extension.. That's all I was pointing out..
it has no relevance, at least no more than using a optical mouse vs a lazor mouse :wtf:

also bear in mind this is on an engineering sample cpu and mobo, not too bad overall at stock clocks
Posted on Reply
#13
w2richwood
it's on first chipset i'm sure there will be more to come
Rich
Posted on Reply
#14
WarEagleAU
Bird of Prey
Its still pretty amazing either way. Records and benchmarks have been using ATI hardware lately. Impressive!
Posted on Reply
#15
EastCoasthandle
Looks like we are seeing diminishing returns. I would have expected a higher score. The Nehalem 2.93GHz is an 8 core cpu right?
Posted on Reply
#16
farlex85
by: EastCoasthandle
Looks like we are seeing diminishing returns. I would have expected a higher score. This is an 8 core cpu right?
No, it's a 4-core cpu w/ hyper-threading, giving it 8 logical cores in window's. I'm not sure what the difference is performance-wise, but I feel like 8 logical cores is probably not the same as 8 real cores.
Posted on Reply
#17
EastCoasthandle
by: farlex85
No, it's a 4-core cpu w/ hyper-threading, giving it 8 logical cores in window's. I'm not sure what the difference is performance-wise, but I feel like 8 logical cores is probably not the same as 8 real cores.
I see, well that would explain the results then.
Posted on Reply
#18
farlex85
by: EastCoasthandle
I see, well that would explain the results then.
It would, really when you think about it, not only is the board and cpu not fully developed, but windows and vantage may be just warming up to the tech new. So many things working together, at this stage in the game I think those scores are just based on raw architecture power, and seem pretty nice.
Posted on Reply
#19
trt740
what socket will N use 775 and if so will it only be for ddr3 based boards?
Posted on Reply
#20
EastCoasthandle
by: farlex85
It would, really when you think about it, not only is the board and cpu not fully developed, but windows and vantage may be just warming up to the tech new. So many things working together, at this stage in the game I think those scores are just based on raw architecture power, and seem pretty nice.
Lets see what develops this time. HT has always been a nice feature, even during the P4 days but it was never really utilized to it's full potential. If we see better support from the OS (Win7), games and other apps things will get very interesting indeed.
Posted on Reply
#21
farlex85
by: trt740
what socket will N use 775 and if so will it only be for ddr3 based boards?
No they got a whole new socket, and the architecture of the chipset has a ddr3 memory controller built in eliminating the fsb, so I'm assuming it's ddr3 exclusive. There was talk at one point of some of the lower models being compatible w/ 775, but that won't be till next year if at all. I don't see how it could w/ the way it operates though. Totally different way of clocking.
Posted on Reply
#22
HTC
by: btarunr
The rather lucky Taiwanese team of Tom's Hardware got their hands on an Intel Bloomfield engineering sample that has a clock-speed of 2.93 GHz, running on a Intel X58 chipset based motherboard made by Foxconn called Renaissance to evaluate a Gainward Radeon HD4850 sample. System details are provided below.

[---]

Of course, the benchmark lacks the advantage NVIDIA PhysX gives to the CPU score in 3DMark Vantage, but for a CPU alone, it is a more than decent score. The system secured P7182 at default settings with a CPU score of 17966. In 3DMark06, it churned out 12786 3DMarks with a CPU score of 5183. In the Crysis CPU benchmark, scores of 33.70 and 18.29 were recorded at 1280x1024 resolution with no anti-aliasing.

Source: Tom's Hardware
Correct me if i'm wrong but i think they're highlighting the CPU score of the respective benchmarks.

Against other quad CPUs @ the same speed, aren't those scores very high?

EDIT

This is, ofc, without the physX thing.
Posted on Reply
#23
btarunr
Editor & Senior Moderator
Yes, the postmortem of my news post is in post #4 in this thread. Dan's 2x 2-core Xeon murders it. :o
Posted on Reply
#24
InnocentCriminal
Resident Grammar Amender
by: EastCoasthandle
Looks like we are seeing diminishing returns. I would have expected a higher score. The Nehalem 2.93GHz is an 8 core cpu right?
Don't forget that the RenaissanceX58 is still in it's infancy, so the BIOS is most likely really immature and hindering the performance of both the CPU and therefore bringing the 4850 down. Could explain why they're running in Dual Channel and not Tri-Channel. So, for an early, initial look at a Nehalem it's pretty promising really. I would really like to see them repeat this test once Foxconn give them a BIOS update, I expect we'd see pretty different results.

As for the 2.93GHz Bloomfield, I thought that this chip was only a quad-core but would handle 8-threads.

Translated version of the original link.
Posted on Reply
#25
lemonadesoda
Nehalem manages between 0% and 40% performance increase clock-for-clock compared to penryn. But with a power consumption cost of 10% for the whole system, which is driven by about 20% extra power for the CPU.

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3326&p=7

So, Nehalem, TODAY, is not impressive at all. You can get the same performance per watt by just overclocking a penryn. However, I'm sure the figures will improve once they optimise mainboard, BIOS, memory channels, and final (non-engineering sample) CPUs will have low power requirements. Or rather, lets hope so, otherwise Nahelem is a flop, and NOTHING like the mammoth win when Intel moved to Core 2 architecture.
Posted on Reply
Add your own comment